CN100525288C - Method and apparatus for large payload distribution in network - Google Patents

Method and apparatus for large payload distribution in network Download PDF

Info

Publication number
CN100525288C
CN100525288C CNB018208142A CN01820814A CN100525288C CN 100525288 C CN100525288 C CN 100525288C CN B018208142 A CNB018208142 A CN B018208142A CN 01820814 A CN01820814 A CN 01820814A CN 100525288 C CN100525288 C CN 100525288C
Authority
CN
China
Prior art keywords
file
content
plurality
server
network
Prior art date
Application number
CNB018208142A
Other languages
Chinese (zh)
Other versions
CN1481635A (en
Inventor
D·C·-H·陈
S·Y·辛
Original Assignee
普里斯梅迪亚网络有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to US26628600P priority Critical
Priority to US60/266,286 priority
Priority to US09/681,644 priority patent/US6970939B2/en
Priority to US09/681,644 priority
Priority to US09/681,665 priority
Priority to US09/681,665 priority patent/US7165095B2/en
Priority to US09/681,672 priority patent/US7047287B2/en
Priority to US09/681,673 priority
Priority to US09/681,669 priority patent/US7181523B2/en
Priority to US09/681,671 priority
Priority to US09/681,669 priority
Priority to US09/681,671 priority patent/US6857012B2/en
Priority to US09/681,672 priority
Priority to US09/681,673 priority patent/US7058014B2/en
Priority to US09/681,668 priority patent/US7177270B2/en
Priority to US09/681,668 priority
Application filed by 普里斯梅迪亚网络有限公司 filed Critical 普里斯梅迪亚网络有限公司
Publication of CN1481635A publication Critical patent/CN1481635A/en
Application granted granted Critical
Publication of CN100525288C publication Critical patent/CN100525288C/en

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network-specific arrangements or communication protocols supporting networked applications
    • H04L67/10Network-specific arrangements or communication protocols supporting networked applications in which an application is distributed across nodes in the network
    • H04L67/1095Network-specific arrangements or communication protocols supporting networked applications in which an application is distributed across nodes in the network for supporting replication or mirroring of data, e.g. scheduling or transport for data synchronisation between network nodes or user terminals or syncML
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10File systems; File servers
    • G06F16/18File system types
    • G06F16/182Distributed file systems
    • GPHYSICS
    • G06COMPUTING; CALCULATING; COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/10File systems; File servers
    • G06F16/18File system types
    • G06F16/188Virtual file systems
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/12Congestion avoidance or recovery
    • H04L47/125Load balancing, e.g. traffic engineering

Abstract

本发明提供网络中大有效负载分布的方法和装置。 The present invention provides methods and devices in the network of large effective load distribution. 本发明的用于对网络中的新节点初始化的方法、系统和装置包括:把新的节点加到网络上,自动地向多个节点发送查询,以确定什么内容需要下载,接收来自具有新节点所需内容的所述多个节点的子集的回答,以及下载来自所述多个节点的子集的内容的所需部分。 Used in the method of the present invention is a new node in the network initialization, system and apparatus comprises: a new node is added to the network, automatically sending a query to a plurality of nodes, to determine what needs to download, having received from the new node answer subset of said plurality of nodes to the desired content, and to download a desired portion of the subset of nodes from said plurality of contents.

Description

网络中大有效负我分布的方法和装置 Large network payload distribution method and apparatus I

技术领域 FIELD

本发明涉及内容交付领域.更具体地,本发明涉及在网络环境中把大有效负栽(即,文件)交付到离用户更近. The present invention relates to the field of content delivery. More particularly, the present invention relates to network environments large effective negative plant (i.e., files) closer to the user to be delivered.

背景技木 Background technology of wood

在网络环境中的内容交付包括把信息(例如,以文件的形式)从内容提供者发送到可以将其内容供应给驻留在两络上的各个目的地的多个用户的多个内容服务器.内容提供者一般把将要被分布的信息放在连接到网络的一个计算机上.这个计算机常常被称作内容服务器.任何客户机-服务器或对等通信协议可以被应用到内容服务器来进一步把信息传送到在相同或不同网络中的被指定来供应信息的一组内容服务器.该源内容服务器通常被称为源服务器.该信息驻留在内容服务器上的一个文件中并且对于网络的用户是可得到的.当用户请求访问该信息时,使用合乎要求的文件传送协议(即,传送的方法)将文件的内容从被指定来供应内容的任何内容服务器交付给请求的用户. 一个内容服务器可以在任何用户请求前从一个源服务器接收信息,或者它可以基于用户请求 Content delivery in a network environment including the transmission information (e.g., in the form of a file) from a content provider to its contents can be supplied to the two network resides on a plurality of users of a plurality of content servers each destination. Usually the content provider information to be distributed on a computer connected to the computer network is often referred to the content server to any client. - server or peer communication protocols may be applied to the content server to further transfer the information to a group of content servers in the same or a different network is designated to supply information. the source content server is typically referred to as the origin server. this information resides in a file on a content server and is available to the user network when a user requests access to this information, using the desirable file transfer protocol (i.e., method of transfer) of the content file delivered to the user to request from any content server is designated to supply the content a content server may be any before the user receives information from a request source server, or it may be based on a user request 一个源服务器检索信息. 一个内容服务器可以被指定来供应来自多个源服务器的信息,并且一个源服务器可以仅仅把其信息的一部分发送到一組内容服务器.内容服务器的所有者通常被 A source server to retrieve information may be a content server specifying information supplied from a plurality of source server and the source server may only send its information to a portion of a set of content servers. Owner of the content servers is usually

称为内容交付网络(CDN)提供者.例如,在诸如因特网的网络中,用户可以经由通过电话公司的中心局(CO)或电缆公司的头端(head end) (HE)连接的因特网服务提供者(ISP)来访问网络.这样,ISP 用作用户到因特网的网关.ISP的例子包括America On LineTM (美国在线)(AOLTM)和EarthlinkTM. —些电话公司和电就公司也是ISP. ISP可以与其它的网络彼此互联,他们可以连接到主千(backbone)提供者、电话公司的网络、电缆^^司的网络或任何私人或公众网络.主干提供者为ISP、企业等提供高的带宽连通性.通过ISP、 CO或HE, 用户可以从网络中的任何内容服务器访问可以从内容提供者得到的服务(例如,数据). Called content delivery network (CDN) provider. For example, in a network such as the Internet, a user may be provided via an Internet connection service through a telephone company's central office (CO) or a cable company head end (head end) (HE) person (ISP) to access the network so that, ISP serves as an example of a user to the Internet gateway .ISP include America on LineTM (AOL) (AOLTM) and EarthlinkTM -.. some telephone companies and electric companies will also be with the ISP ISP. other networks interconnected with each other, they can be connected to the backbones (backbone) providers, the telephone company's network, cable network ^^ Division or any private or public network backbone providers to provide high-bandwidth connectivity to ISP, enterprise, etc. through the ISP, CO, or HE, the user may provide service (e.g., data) are obtained from any of the content from the content server in the visited network.

各种类型的数据(即,信息)可以在一个网络上传输.例如,当一个用户期望访问可以从远程内容服务器得到的网页、文本文档、应用程序、静态困象、音頻、视頻或任何其他类型的数据时,則包含合乎要求的数据(即,信息)的文件内容必须从内容服务器交付给用户. 与一些其它文件类型、例如包含視頻或多媒体数据的文件相比,包含网頁和文本文档的文件一般较小.因此,从位于远程位置、例如澳大 Various types of data (i.e., information) may be transmitted over a network. For example, when the user desires to access a web page from the remote server to obtain the content, text documents, application programs, static trapped as, audio, video or any other type when the data is in line with the requirements of data (ie, information) of the contents of the documents must be delivered from the content server to the user are included. some other file types, such as video or multimedia files containing data compared to contain web pages and text documents files are generally small. Therefore, from a remote location, such as Australia

利亚的内容服务器传送一个网页到美国的用户可以花费不到几秒钟. 但是,例如,根据視频文件的大小和用户连接的速度,传送視频文件可能花费数分钟到数小时.这样的传送在网络上放置了大量的需求, 可能会导致丢失数据.例如,当数据通过因特网发送时,接收系统可能接收不到从内容服务器发送的所有数据.这是因为数据包(数据一般以包来传送)可能通过一些路由器,在这里一些包由于拥桥被丢掉. 接收系统通知服务器所丢失的数据,使得它可以重发该数据.在一些情况中,丢掉的包可以延緩或暂停内容的交付,因为如果许多服务器把重发的数据保存到其客户机,則路由器变得更拥挤,这样导致更多的被丢掉的包. Leah content server transmits a web page to users in the US can cost less than a few seconds. But, for example, depending on the speed of the video file size, and user connections, transfer video files may take a few minutes to several hours. Such transmitting a large number of requirements placed on the network may cause the loss of data. for example, when data is transmitted via the Internet, the receiving system may not receive all data transmitted from the content server. this is because the packet (data packet to generally transfer) may be through some routers where some packets are dropped due owned bridge receiving system notifies the server of the missing data so that it can retransmit the data. in some cases, dropped packets can be delayed or paused content delivery, because many servers saved if the retransmitted data to its client, the router becomes more crowded, this has led to more packages being lost.

根振被传送的文件的大小,依賴一个单独的源同时把各种类型的信息分布到多个远程位里的基于网络的内容交付可能遇到有关服务器的网络加栽问题或服务器本身可能被过度分派任务.例如,由于传送一个小文件(例如, 一个网页)通常仅仅花费几秒钟,所以小文件从一个源位置到成千上万的目的位置的大量分布对源位置附近的网络交通不会产生大的影响.相反,传送一个大文件(即, 一个大有效负栽) 可以花费数十分钟到若干小时.如果这样的有效负栽的分布依赖于一个单独的源,則源位置附近的网络性能和随后的内容的交付会严重地降低并且变得不能接受. The size of the root vibration transmitted documents, rely on a single source while the distribution of various types of information related to the delivery server's network may encounter problems plus plant or server-based content in a more remote location itself may be excessive assigned tasks. for example, due to the transfer of a small file (for example, a web page) typically takes only a few seconds, so a large number of small files from a distribution source location to the destination of thousands of network traffic in the vicinity of the source location will not a big impact. in contrast, a large file transfer (i.e., a large payload plant) can take several tens of minutes to several hours. If such payload depends on the distribution of plant a single source, near the location of the source network delivery performance and subsequent content will be severely reduced and becomes unacceptable.

因此,虽然依賴于一个单独的源分布小文件(例如,网页、文本或小困象)是可以被接受的,但服务器和/或网络过栽的潜在可能要求使用多个源来分布大文件到多个客户机, Thus, although a single source depends on the distribution of small files (e.g., web pages, text, or small storm image) it is acceptable, but the server and / or network planted over the potential requires multiple sources to distribute large files to multiple clients,

宽带工业的快速发展已经推动了丰富媒体(rich media)(例如, 完全长度的电影、視頻或其它类型的多媒体数据)的发展.宽带技术带来了将内容交付给远程用户的高速连接的能力,闳此大有效负栽可以传送得更快.并且,宽带技术使利用流型(streaming)媒体发送音频和/或視频数据成为可能一例如,由此数据以流的方式被发送以供实时重放.这样,在用户终端上的丰富媒体的质重,现在比任何其它类型的信息的质量更依赖于交付技术的执行能力.为了最小化交付延迟、两络拥挤和其它相关问趙, 一些系统试田将内容放在位于近距离接近、即距离最终用户几个网络集线器远的服务器系统上.这些服务器位置近似地定义了通常所说的网络的"边缘"的概念.例如,因特网服务提供者近距离接近最终用户,这样可以被看作处于网络的边缘.当服务器被置于这样的位置时, The rapid development of broadband industry has driven rich media (rich media) (eg, full-length movies, video, or other types of multimedia data) the development of broadband technology offers the ability to connect to high-speed content delivery to a remote user, this plant Hong large payload can be transmitted faster. Also, broadband technology makes use of the flow pattern (streaming) media transmitting audio and / or video data of a possible example, this data stream to be transmitted for real-time manner heavy put. Thus, rich media at the user terminal, quality is more important, now more than quality information of any other type are more dependent on the ability to execute delivery techniques. in order to minimize delivery delays, two network congestion, and other related Q. Zhao, some systems the content in the test field is located in close proximity to, i.e. these server location approximately define the concept of "edge" of the network known as the end-user from the server system several hub away. For example, an Internet service provider close proximity to the end user, which can be seen at the edge of the network when the server is placed in such a position, 服务器被称为处于网络的边缘. 被配置为从位于网络边缘的网络节点获得内容的最终用户系统因此在 It is referred to as edge servers in the network. The system is configured for the end user to obtain content from a network node in the edge of the network and therefore

网络边缘(也叫作最近距离(lastmile))之外.但是,注意到位于网络边缘之外的系统仍然连接到网络并且能与位于边缘上的服务器计算机进行通信是很重要的.在网络边緣放置内容是有利的,因为它可以减少在服务于位于边緣之外的用户时的延迟.把大的有效负栽交付到"边缘"的当前方法包括镜象法(mirroring)或超高速緩存法(caching).现在将详细讨论这些方法和每一种方法中固有的局限性, 以便让读者了解本发明所产生的进步. Network edge (also called the closest distance (lastmile)) outside. However, it is noted outside the edge of the network system is still connected to the network and is capable of communicating with the server computer important located on the edge of the network edge placing content is advantageous because it can reduce the delay in serving the user is located outside of the edge. the large payload delivered to the plant "edge" of the current methods include mirror Act (mirroring) or cache method (Caching). these methods will now be discussed in detail and the limitations inherent in each method to allow the reader to understand the progress of the present invention is produced.

超高速緩存法 Cache method

超高速緩存的一个筒单的例子是网络超离速緩存.以它的袭简单的形式,两络超高速援存涉及位于客户机用户和源服务器之间的超高速緩存器件,使得曾经从源服务器取出的数振被保存在超高速援存设备(器件)以服务于随后的对于相同数振的请求.例如,超高速援存的示例显示在困1中.在局域网(LAN) 108中的浏览器104上的期望获得可以从源服务器100得到的数据的客户机用户把合乎要求的数椐的通用资源定位器(URL)地址榆入到浏览器104中,例如,LAN 108 可以是一个ISP的两络.该请求被转发到超高速緩存器件102,它在这个示例中是一个HTTP (超文本传送协议)代理服务器.例如,可以由ISP所拥有的代理服务器典型地位于ISP的本地网络,象任何其它服务器一样,代理服务器(超高速緩存器件)102和IO3是具有本地处理和存储器的计算机.那个存储器的一个子集合被称 Examples of a single cylinder of the cache from the cache over a network. In its simplest form the passage, two ultra-high speed recovery network storage device relates to cache located between the client and the origin server user, from a source that was remove the server the number of vibration are stored in ultra-high speed recovery storage device (the device) e.g., exemplary ultrafast recovery stored display in a local area network (LAN) 108 in storm 1 to service subsequent requests for the same number of vibration. desired on the browser 104 is obtained from the client user data obtained by the source server 100 desirable number noted in the universal resource locator (URL) address Yu into the browser 104, e.g., LAN 108 may be an ISP two network. the request is forwarded to cache device 102, which is a HTTP (hypertext transfer protocol) proxy server in this example. For example, typically located at the ISP's local network by the proxy server ISP owned, like any other servers, proxy servers (cache device) having a computer 102 and IO3 local processing and memory. the memory is a subset of the known 为代理超高速緩存.超高速緩存一般用作頻繁使用的信息的临时存储器.注意,虽然仅仅一个超高速緩存器件显示在图1的每一个ISP的局域网中,但是实际的实施可以在一个ISP的局域网中具有多于一个的超高速緩存器件. A proxy cache. Usually used as a temporary cache memory of the information frequently used. Note that, although only one cache devices shown in each ISP's local network of FIG. 1, but the actual implementation may be an ISP LAN has more than one cache device.

代理服务器(即,超高速緩存器件)102处理在浏览器104上从客户机接收的请求并且从它的超高速緩存(即,存储器)技索被汆求的数据,如果该数据不能从它的超高速援存得到,代理服务器102经由网络路由器101把该请求转发到源服务器100.在这个示例中,网络路由器101唯一的用途就是把请求转发到源脹务器100,源服务器100是具有到在樹览器104的客户机用户的单个TCP/IP(传榆控制协议/因特网协议)连接路径110的HTTP服务器. Proxy server (i.e., the cache device) 102 processing request on a browser 104 receives from the client and from its cache (i.e., memory) technology cable is boil requested data, if the data is not available from its ultra high-speed memory to obtain aid, proxy server 102 via network router 101 forwards the request to the origin server 100. in this example, the network router 101 only purpose is to forward requests to a service expansion 100, having the source server 100 single TCP client user in the tree view 104 / IP (transmission elm control protocol / Internet protocol) connection path 110 to the HTTP server.

源服务器100服务于该请求并且把所请求的数据转发到超高速緩存器件102.在接收到数据之后,超高速緩存器件102可以把数据保存到它的本地超髙速緩存存储器中并且也把它转发到浏览器104.数据据说被超高速緩存在HTTP代理(超高速援存器件)102中.在浏览器105的期望相同数据的随后的客户机用户得到由HTTP代理服务器(超高速緩存器件)102服务于他们的请求而不需要将该请求转发到HTTP 服务器100.但是,在LAN109上请求相同数据的用户106和107将由HTTP服务器IOO来服务于他们的初始请求,因为用户106和107没有通过具有超高速緩存于存储器中的数据的HTTP代理102连接.而是HTTP代理103将执行与上面对于HTTP代理102所讨论的一样的处 The source server 100 to service the request and forwards the requested data to cache device 102. After receiving the data, the cache device 102 can save the data to its local cache memory ultra Gao and it also forwarded to the browser 104. the data is said to be in the cache HTTP proxy (ultra high speed storage device DAC) 102 was obtained (cache device) by the HTTP proxy server in subsequent browser client user data 105 of the same desired 102 serve their request regardless forwards the request to HTTP server 100. However, the same data is requested on the user LAN109 106 and 107 by the HTTP server IOO to serve their initial request, because the user does not pass 106 and 107 HTTP proxy cache 102 has a data memory in connection instead HTTP proxy 103 performs the same as above for HTTP proxy 102 discussed at

理来获得数据并把该数据超高速援存到它的存储器中.这样,所述的在网络边缘的代理服务器102和103根据用户的需求而分布, Processing to obtain the data and sends the data stored to aid its super high speed memory. Thus, the edge of the network 103 and the proxy server 102 according to user's demand distribution,

一旦数据被超高速緩存到HTTP代理102和103中,源服务器100 将不需要服务于来自通过HTTP代理服务器102和103连接的用户对于相同的数据的请求.通过在更靠近用户的各个代理服务器上超高速援存数据,来分布内容的交付,从而减少网络服务器周围的负栽.但是,超高速緩存仅仅对于交付静态内容一在存储器中是固定的数据、 例如静态网页是有益的.超高速緩存对于动态信息、例如服务(例如, 功能、交易等)、流型媒体或其它任何类型的动态信息是无效的. Once the data is cached to the HTTP proxy 102 and 103, origin server 100 would not need to service the HTTP proxy server 102 from the user 103 and the connection request for the same data by the proxy server on each closer to the user ultra high-speed DAC deposit data to the distribution of content delivery, thereby reducing the negative planted around the network server. However, the cache only for the delivery of static content a fixed data in memory, such as static web page is useful. cache for dynamic information such as services (e.g., functions, transactions, etc.), streaming media type or any other type of dynamic information is invalid.

HTTP协议对于本领域中的普通技术人员是公知的;因此在HTTP 代理服务器102和103上执行超高速緩存功能的软件是容易得到的. 但是,这对于流型媒体不适用,闳为流型服务器的不同的提供者使用不同的协议来传输数据到接收者播放器(例如,浏览器),图2是一个到播放器的典型的流型服务器连接的示例.与到對览器的HTTP TCP/IP连接对比,流型服务器200通过三条连接路径连接到播放器201.路径202是实时流型协议(RTSP)连接. RTSP是为控制具有实时特性的数据例如音頻和視频流的交付而提供的一个协议.RTSP包皿体数据的描述并为播放器201提供例如播放、倒带、快进和暂停的重放控制.重放可以以一个偏移完成,使得播放器可以从一个規定点开始接收数据.例如,当播放器201倒带时, 对应于合乎要求的重放位置的一个不同的偏移被发送到流型服务器200,并且输入的 HTTP protocol of ordinary skill in the art is well known; thus executing software cache function on HTTP proxy servers 102 and 103 are readily available, however, that for streaming type medium is not applicable, Xiahong flow pattern server. different providers use different protocols to transfer data to the recipient player (e.g., a browser), FIG. 2 is an example of a typical flow pattern server connected to the player. of the browser to the HTTP TCP / Compare IP connection, flow pattern 200 is connected to the player server 201. the path 202 is a real-time streaming protocol (RTSP) is connected via three connection paths. RTSP is provided to control data having a real-time characteristics such as delivery of audio and video streams provided a packet protocol .RTSP dishes as described in the volume data and provides the player such as play, rewind, fast forward and pause playback control 201 can be accomplished in a shift playback, so that the player can start from a predetermined receiving point data. For example, when player 201 rewinds, the reproduction position corresponding to the desirable offset of a different type is transmitted to the streaming server 200, and the input 据从一个新的偏移开始通过路径203被发送.路径203利用实时传送协议(RTP )并且可以包含被重放的数据.第三个连接、路径204利用RTP控制协议(RTCP)并且它可以提供数据的流程控制. It starts from a new 203 is transmitted through the shift path 203 path using real-time transport protocol (RTP) and may contain data to be reproduced. The third connection, path 204 using RTP Control Protocol (RTCP) and it may be provided controlling the flow of data.

超高速緩存对于流型媒体效果不好,因为流型服务器的各个提供者使用不同的智能来计算作为偏移和流程控制的函数的在连接203上发送的数据.而且,厫务器提供者不遵循公用标准,因此把超高速援存器件放置在流型服务審200和播放器201之间将不是容易可行的,除非在现在的实施中是在流型服务器中的智能被包括在连接路径上发送的信息流中,或者如果超高速緩存器件包含由每一个流型服务器提供者使用的智能.这样,现有的系统目前没有提供一个可行的方法来超高速緩存流型媒体数据.并且,由于超高速諼存是基于使用的,当内容没有被超高速緩存时,代理将需要取该内容,因此存在丢失的可能并且没有质量的保证, Cache for streaming media type not good, because the individual providers use different flow patterns of the server calculated as a function of the intelligent flow control and the offset of the data transmitted over the connection 203. Also, Ao service provided does not follow standard public, so the ultra high-speed recovery storage device disposed between the flow-based service 200 and the player 201 trial will not be readily feasible unless the intelligence is in flow pattern server is included in the connection path in the present embodiment the stream sent, or if the cache contains intelligent device to use a flow pattern by each server. Thus, the conventional systems do not currently provide a viable way to cache streaming media data type. Further, since the Xuan is based on ultra-fast memory used when no content is cached, the proxy will need to take the content, so there may be lost and there is no guarantee of quality,

尽管这些局限性,超高速緩存法仍具有许多优势,例如容易扩充, 因为新的超高速緩存器件可以被添加到任何地方并且它将启动并运行;超高速緩存器件可以被不同的内容提供者共享;并且超高速緩存器件非常轻便(即,不需要特別的配置),这样更容易管理. Despite these limitations, the cache method still has many advantages, such as easy to expand, because the new cache devices can be added to any place and it will be up and running; cache devices can be shared by different content providers ; and cache devices is very lightweight (i.e., does not require special configuration), it is easier to manage.

镜象法是用于为在网络"边缘"的用户提供内容交付的方案,该方案通过将内容复制到两络边缘、从而使内容被请求的地方和内容被服务的地方之间的距离最小化,解决了集中式系统的许多局限性.在这样傲时,与从一个集中式源位置交付到多个用户相比,镜象法节省了网络带宽.镜象法的基本原理包括内容和网络的中心控制、内容到网络边缘上的服务器的有效分布和从用户到本地边缘服务器的内容请求的自动转移. Mirror method for providing content delivery scheme for users in a network "edge", the contents of the program by copying the edge network to the two, so that the requested contents and the local distance between the place was minimized and services address many of the limitations of centralized systems. in so proud, as compared with delivery from a centralized source to a plurality of user positions, mirror method saves the network bandwidth. the basic principle of the method of images including content and network central control, efficient distribution of content to a server on the edge of the network to the local user and the content from edge server automatically transfer request.

在镜象法中,文件服务器被遍及网络(例如罔特网)放置,靠近内容请求起源的地方.这个原理反映了超高速緩存的一些功能,但是具有明显的差別.特别地,这些文件服务器以中心控制的协作方式一起工作以确保整个网络的性能.象超高速緩存一样,内容从源服务器到该服务器仅仅被复制一次,而不考虑内容被供应的次数.但是,镜象法提供了更大的内容控制.通过预填充服务器,内容将可得到以供快速交付给用户,消除超高速緩存丢失并且增加命中率.镜象法与超高速緩存法结合,提供了一种具有两种方法的好处的更好地综合的结果. In the method of images, the file server is across the network (for example, ignored the special net) placed near the origin of the contents of the request. This principle reflects some features of the cache, but with significant differences. In particular, these file server work Center control ways of collaborating together to ensure the performance of the entire network. like the same cache, once content is copied from the source server to the server only, regardless of the number of times content is supplied. However, the mirror method provides greater the content control by pre-populating the server, the content will be available for quick delivery to the user, eliminating the cache hit rate and increase the loss. mirroring method combined with the cache method provides the benefit of having two methods consolidated results better.

一个URL应用到一个镜象法实施方式中的所有服务器.当浏览器请求URL时,系统基于如下几点确定一个本地交付服务器:地理和网络位置;内容的存在;和服务器的当前状态(可用性和负栽). A server URL applied to all embodiments of a method when mirror browser request URL, the system determines the following points based on a local delivery server: geographical and network location; content exists; and a current state of the server (and usability negative plant).

图3是采用镜象法把内容推到网络边緣的网络内容交付方案的困示.如困3中所示,假设边界300表示网络的边缘,镜象法将文件服务器(例如,FS 301-308)定位在边缘.在这个示例中,文件服务器301是控制所有其他文件服务器(例如,302-308)的主服务器.需要被推进到边缘的所有内容被加栽到主服务器301中,然后使用最佳的推进方法复制到所有的其他文件服务器302-308中.例如,内容可以使 Figure 3 is the mirror image of the content push method to the edge network content delivery network programs trapped shown. As shown three trapped assumed boundary 300 represents the edge of the network, the method mirror file servers (e.g., FS 301- 308) positioned at an edge. in this example, the file server 301 is to control all other file servers (e.g., 302-308) of the master server needs to be advanced to the edge of the content is applied to all the main server 301 plummeted, then copy the best method to promote all of the other file servers 302-308. For example, content can be made

用下面讨论的多点传送来复制. Multicast to copy with the following discussion.

不象超高速緩存法,内容必须是静态的(即,不随时间而改变), 镜象法对于诸如交易的非静态数据效果好,因为交易数据从主服务器(例如,FS301)到在网络边缘的文件服务器(例如,FS302-308)可 Unlike cache method, the content must be static (i.e., does not change with time), mirror image method for non-static data such as transactions good results, because the transaction data from the master server (e.g., FS301) to the edge of the network file server (for example, FS302-308) can

以同步,复制数据到边缘上的文件服务器的各种方法可以包括广播, 从主服务器到网络中的所有收听的文件服务器的传输;任意点传送(anycast),到最近的服务器组的传输;单点传送(unicast),到一个特定接收器的传输;以及,多点传送,到多个特定接收器的传输(多点传送的更详细的讨论将在下面讨论). 一旦内容在边缘被交付,在浏览器330的请求访问内容的用户自动地传递到地理上最近的能够服务那个请求的服务器(例如,服务器307). To synchronize, copy the data to a variety of file server on the edge may include broadcast, a transmission from the master server to listen to all the network's file server; anycast (Anycast), a transmission to the nearest group of servers; single transfer point (unicast), to a specific transmission receiver; and a transmission, multicast, to a plurality of specific receivers (a more detailed discussion of multicasting is discussed below) Once the content is delivered at the edge. user requests access to content in the browser 330 is automatically transmitted to the server geographically closest that can service requests (eg, server 307).

镜象法对于流型媒体效果也很好.流型服务器可以被连接到任何一个文件服务器301-308来为最接近需要的地方提供服务.例如,通过将流型泉务器310连接到文件服务器302,在地理上邻近文件服务器302 的播放器320上的用户可以重放流型媒体数据而没有太多的延迟.这样,在镜象法实施方式中,流型服务器可以被连接到任意的文件服务器以克J8SJ^高速緩存法的局限性.但是,当前的方法也有显著的缺点, 例如,因为对内容的重复访问以及因为盘输入/输出带宽是有限的,大的目标、例如常见的視频可能在盘上产生热点.况且,在对于数据的最终用户客户机请求开始满意之前,大的目标需要被完全地传送到应用服务器或超高速援存器件,从而产生潜在的延迟问趙. Mirror method is also very good results for streaming media type. Flow server may be connected to any of the file servers 301-308 to provide service closest to where needed. For example, flow pattern 310 is connected by springs to the file server traffic is delay 302, in the geographic proximity user on the file server 302 may be player 320 reproducing media data flow pattern not much. Thus, in the embodiment, the mirror method, the flow pattern can be connected to any server in the file server in limitations g J8SJ ^ cache method. However, the current methods have significant drawbacks, e.g., as repeated access to the content, and the disk input / output bandwidth is limited, a large target because, for example, the common video may hot spots on the disc. Moreover, prior to the end user client data request has been happy, big goals need to be fully transmitted to the application server or ultra-high-speed memory devices aid, resulting in potential delays asked Chao.

镜象法也会由于可伸缩性问題、存储器限制、管理成本和不适当的负栽均衡而非常昂责.可伸缩性问趙起因于在一个存储媒体中存储整个大文件、例如视頻的需要.因此,当可用的存储器不足以存储特別大的文件时,新的存储器必须被添加到网络中的所有文件服务器.因为网络中所有的文件歉务器必须保持相同的文件配置,在镜象法环境中升级所有的文件服务器可以证明是非常昂贵的.另外,带入到网络 Since the law will mirror scalability issues, memory constraints, management costs and negative inadequate planted balanced and very expensive responsibility. Scalability asked Zhao attributed to store an entire large files in a storage media, such as video needs. Thus, when the available memory is insufficient to store extremely large files, new storage must be added to all the file servers in the network. since all of the files in the network traffic apology must maintain the same file configuration, process environment Mirror upgrade all of the file servers can prove to be very expensive. Also, bring to the network

中的新的文件服务器将需要被配置成与网络中的所有其他文件服务器一致. The new file server will need to be configured to be consistent with all other file servers on the network.

添加更多的存储器需要用于安装新的存储设备的机架空间.机架空间通常是有限的并且有时是昂责的.况且,随着存储器容量增加,需要更多的系统管理功能(例如,备份)来管理所述配置.由于系统管理的成本昂贵并且机架空间有限,所以镜象法受到影响. Adding more storage requires rack space for mounting the new storage devices. Rack space is usually limited and sometimes expensive liability in. Moreover, as the memory capacity increases, more system management functions (e.g., backup) to manage the configuration. expensive due to the cost of system management and rack space is limited, so the method of images is affected.

使用多点传送的内容分布 Use the multicast content distribution

多点传送是在两格上的单个发送者和多个选定的接收者之间的同步通信.困4是使用多点传送技术把信息推进到网络上的多个内容服 Multicast is communication between a synchronous sender on both single cells and a plurality of selected recipients. 4 is trapped using multicast technology to propel a plurality of content information to a server on a network

务器的一个分布网络的示例. An example of a business's distribution network.

源提供者把大的有效负栽(例如,视頻文件、困象数据或具有足以使网络资源紧张的尺寸的其它任何文件)上栽到例如可以是一个位于Los Angeles的内容風务器的根服务器400上.根服务器也可以被称为源徵务器.接着根服务器400把視頻数据多点传送到处于网络服务器树的第二級上的通常在不同的地理位置的多个服务器(例如,服务器柳1到403) •例如,厫务器401可以位于San Diego,服务器402位于San Jose,服务器403位于San Francisco.在接收视频数据后,服务器401到4紛将把視频數据多点传送到服务器树的下一级中的服务器.例如,服务器401把这些数据多点传送到服务器404到406,服务器402 把这些数据多点传送到服务器407到409,服务器403把这些数据多点传送到菔务器410到412.在这个示例中,每一个服务器多点传送到三个其他服务器,但是,大多数实施涉及到多于三个服务器 The source provider planted large payload (e.g., video files, image data, or any other difficulties file having sufficient dimensions strain on network resources), for example, may be planted to a content located in Los Angeles air traffic in the root server 400. the root server may also be referred to as source traffic sign device. root server 400 then video data is typically multicast to a plurality of servers (e.g., servers on different locations on the network server in the second stage of the tree Liu. 1 to 403) • For example, works Ao 401 may be located in San Diego, server 402 is located in San Jose, server 403 is located in San Francisco. upon receiving the video data, the server 401-4 to the divergent will multicast the video data server under a tree of the server. For example, these data server 401 to the server 404 to multicast 406, multicast server 402 to the data server 407 to 409, the server 403 to multicast data traffic turnip 410 to 412. in this example, each of a multicast server to three other servers, however, most embodiments relate to more than three servers (例如,十个服务器)的多点传送. (E.g., ten servers) multicast.

在裯,頻数据被分布到服务器400到412中之后,视頻数据变成可以从位于网络上的不同的地理位置的多个服务器得到.这个分布方法把内容推进到边缘成为镜象类型的体系结构,在这里可以从多个服务器中的一个、通常从地理位置最近的服务器来服务于用户的请求.由于 After the coverlet, the audio data is distributed to the server 400 to 412, the video data becomes available from multiple servers on different geographic locations located on the network. The content distribution method to be advanced to the edge of the mirror type of architecture , from where a plurality of servers, usually from the nearest location server to service the user's request. Since

在一特定通信链路上不足的容量,多点传送整个大有效负栽文件仍然可以造成拥挤;由于网络设备的处理速度而造成网络设备拥挤;由于服务器的数椐处理速度而造成服务器拥挤;以及由于与数据在长距离上传送有关的时间而造成的网络中的延迟. Insufficient capacity on a particular communication link, multicast entire large payload file may still cause congestion plant; since the processing speed of the devices network congestion caused by a network device; as noted in the number of the processing speed of the server causing the server is congested; and Since the network-related data transmission over long distances in the time delay caused.

负栽均衡 Negative balanced plant

负我均衡是分布网络负栽和处理到一簇服务器的负栽以提高系统性能、同时同步增加由服务器提供的服务的可靠性的任务.负栽均衡器经常被实现为开关或路由器并被分別称为负栽均衡开关或负栽均衡路由器.负栽均衡器的网络接口、虚拟IP地址(V1P)用作脹务器簇的虚拟外部接口.簇中的每一个厥务器具有一个内部(本地IP地址) 和一个外部(1P地址)网*^口.大多数负栽均衡器提供被称为网络地址翻译(NAT)的特征,它把VIP翻译成因特网上可用的本地IP地址.负栽均衡器接受所有寻址到其VIP的数据包,并将其均匀分布到最可得到的服务器. I negative negative equilibrium is distributed network processing plant and to plant a cluster of servers to improve system performance negatively, while simultaneously increasing the reliability of task services provided by a server. Planted negative equalizers is implemented as a switch or router, respectively, and called negative negative plant or plant balancing switch balancing router. negative plant equalizer network interface, a virtual IP address (V1P) serves as a virtual external interface service expansion device cluster. each cluster having an internal task Jue (local IP address) and an external (1P address) network * ^ mouth. most of the negative plant equalizer provides feature is called network address translation (NAT), which translated into the local IP address of the VIP available on the Internet. negative balanced plant accepts all addressed to the VIP packets and uniformly distributed to the most available servers.

负栽均衡器保留一个状态表(例如,什么服务器正在服务于什么用户),使得持夂对话期的数据包流向或来自相同的客户机或服务器终点.许多负栽均衡器具有把数据包从客户机分布到客户机先前被连接到的相同的服务器的可配置的"粘性"特征.此"粘性"特征允许服务器智能地为来自其客户机的可能的未来请求作准备. Negative planted equalizer maintains a state table (for example, what the server is to serve what the user) so that packets pass through holding Wen dialogue period or from the same client or server end. Many negative planted equalizer with a data packet from the client machine distributed to the same server may be configured client was previously connected to the "sticky" feature. this "sticky" feature allows a server to intelligently possible future requests from its clients to prepare.

负栽均衡器可以典型地以"常规"(即,不透明)模式或"透明"模式运行."常规"棋式和"透明"模式之间的区別在于入站和出站数据流的管理.在"常规"模式中,所有到服务器簇的入站通信量和来自服务器簇的出站邇信量都通过负栽均衡器.在"透明"模式中,来自服务器簇的出站通信量通过直接流过IP路由器来绕过负栽均衡器."透明"模式对于交付大量数据的服务器的网络可以是非常重要的,因为它减少了在负栽均衡路由器上的总的负栽并西而提高了网络的性能.当负栽均衡器以"透明"模式运行时,它不翻译在从客户机到它的服务器组的入站包中的目的IP. 1P路由器必须被同时连接到负栽均衡器和服务器簇来完成这一点.然后菔务器簇中的服务器用使用负栽均衡器的IP地址的回送接口和用到IP路由器的缺省路由来配置. Negative plant equalizer may typically operate in "normal" (i.e., non-transparent) mode or a "transparent" mode. "Normal" chess problems and "transparent" mode lies in the management of the difference between the inbound and outbound data streams. In "normal" mode, all inbound traffic to the server cluster, and the amount of outbound near channel from the server cluster through a negative plant equalizer. in "transparent" mode, traffic is outbound from the server clusters by direct flow over-IP router to bypass the negative planted equalizer. "transparent" mode can be very important for the network server to deliver large amounts of data, because it reduces the overall negative in the negative planted and planted west of balancing router improves network performance when negative plant equalizer operating in "transparent" mode, it does not translate the destination IP in from a client to its server group inbound packet. 1P router must be simultaneously connected to the negative planted equalizer and server cluster to complete it. then turnip service is cluster server by using a negative planted equalizer IP address of the loopback interface and IP router uses the default route to configure.

大多数负栽均衡審提供远程或本地应用程序设计接口(API)或者脚本来管理其负栽均衡任务. 一般地,当前技术使用循环法(即,队列中的下一个服务器服务下一个客户机)来负栽均衡一簇可用的服务器.这可能意味着即使服务器没有可用的带宽它们也被分派了任务. Most negative balancing trial planted to provide local or remote application programming interface (API) or scripts to manage the negative planted balancing task. In general, the use of the current technology cycle method (ie, the queue at a server under a service client) to plant negative balanced cluster server available. this could mean that even if the server is not available bandwidth they are also assigned the task.

因此,需要解决与向网络边緣交付大有效负栽相联系的成本、可伸縮性和负栽均衡问趙.但是,在讨论本发明之前,展示了如何在不同的操作系统中处理文件的总体概述. Therefore, the need to address delivered to the network edge with large negative cost effective plant linked, scalability and negative planted balanced asked Zhao. However, before discussing the present invention, showing overall how to handle files in different operating systems Overview.

计算机系统的文件配置 Computer system configuration file

在操作系统中文件被命名、存储、组织和访问的全部结构被称为"文件系统".例如,在UNIX操作系统中,每一个目录都可以用一个文件系统安装.如果一个目录/X用文件系统Y安装,则在子树/X中的任何存储器I/O (输入/输出)请求被转发到文件系统Y.例如,文件/X/foo.txt的打开使得打开请求被转发到文件系统Y中相应的"打开"例程. Files are named in the operating system, the entire structure of the storage, organization and access is called a "file system". For example, the UNIX operating systems, each directory can be mounted with a file system. If a directory / X file with installation system Y, any subtree memory I / X of the / O (input / output) request is forwarded to the file system Y. For example, opening a file that /X/foo.txt open request is forwarded to the file system Y in the corresponding "open" routine.

现在的操作系统,例如Unix和Windows,支持"可堆叠文件系统". Current operating systems, such as Unix and Windows, support for "stackable file system."

可堆叠文件系统是建立在另一个文件系统顶部的一个文件系统.例如,如果可堆叠文件系统F在文件系统K之上建立,并且如果目录/X用F安装,那么文件/X/foo.txt的打开使得打开请求转到文件系统F,文件系统F处理该请求,它可以产生或可以不产生到文件系统K的请求.在Windows操作系统环境中,可堆叠文件系统被称为文件筛选程序(filter).文件筛选程序可以被放置在任何目录上.对具有文件筛选程序的目录的任何I/O访问使得相应的文件筛选程序例程将被执行.文件筛选程序可以发送或可以不发送任何请求到下层的文件系统. Stackable file system is to create a file system on top of another file system. For example, if a stackable file system F is built on a file system K, and if directory / X is mounted with F, then the file /X/foo.txt open request to the file system so that the open F, file system F processes the request, it may create or may not generate a request to file system K. in the Windows operating system environment, a stackable file system is called a file selection procedure ( filter). file screening procedure may be placed on any directory. directory has files to any I selection procedures / O accesses so that the corresponding file filter routine to be executed. screening can send files or may not send any request to the underlying file system.

被分布的文件系统是其中文件可以位于连接到局城和广域网上的多个服务器上的系统.被分布的文件系统可以使用几个公知的网络文件系统协议,例如通用闳特网文件系统(CIFS )和Sun Microsystems公司的网络文件系统(NFS)协议来实施.CIFS是基于由运行各种各样的操作系统的个人计算机和工作站广泛使用的标准服务器消息块(SMB)协议的.CIFS协议支持多个文件共享和表示特性,例如:文 The distributed file system in which file system may be located on multiple servers connected to the local city and the wide area network. The distributed file system may use several well-known network file system protocols, such as Universal Hong Laid Network File System (CIFS .CIFS protocol) and Sun Microsystems, Inc.'s network file system (NFS) protocol is based on the standard implemented .CIFS server message block by the operating system to run various personal computers and workstations widely used (SMB) protocol to support multiple file sharing and a characteristic, such as: text

件访问、文件和记录铺定、安全的超高速緩存、预读和后写、文件改变通知、协议版本协商、扩充属性、分布的复制虚拟巻和服务器名称分解.象CIFS—样,NFS试闺为客户机系统提供一个开放的交叉平台 Piece access, file and record shop set, safe caching, read-ahead and write, file change notification protocol version negotiation, extended attributes, copy distributed virtual server name Volume and decomposition. CIFS- kind of like, NFS test boudoir the client system provides an open, cross-platform

机制,以便在网络上从服务器系统请求文件服务.NFS协议提供在网络中对于共享文件的透明远程访问,因为它被设计为可在不同机器、操作系统、网络结构和透明协议之间移动.NFS的可移动性通过在使用外部数据表示标准(XDR)的系统实现的顶部建立的远程过程调用基元(primitive) ( RPC基元)的使用来实现.RPC基元提供到远程服务的接口.服务器提供程序(例如,NFS),每一个程序包括一组过程.服务器的网络地址、程序号和过程号的组合指定一个要执行的具体的远程过程.XDR使用语言来描述数据格式.语言只能用于描迷数据,它不是一个程序设计语言.NFS实现存在于各种各样的系统中.NFS安装协议允许服务器将远程访问特权分发给一个受限制的客户机集合,并执行例如允许将一个远程目录树连到一个本地文件系统的各种操作系统特定功能. Mechanism, so as to request from the server system on a network file service .NFS protocol provides transparent remote access to shared files on the network, since it is designed to be moved between different machines, operating systems, network protocols, and the transparent .NFS the top may be used to establish a remote system using the process external data representation standard (XDR) implemented call primitives (primitive) (RPC primitives) to achieve .RPC primitives provided to the remote service via the interface mobility server provider (e.g., the NFS), each program comprising a set of process network address of the server, the program number and combination of processes specified number of a specific remote procedure to be executed .XDR language to describe data formats. language only with fans to describe the data, it is not a programming language .NFS implement agreements exist in a wide variety of installation systems .NFS allows the server to be distributed to remote access privileges to a client machine collection restricted, and for example allows a remote execution directory tree attached to a variety of operating system-specific functions to a local file system.

上述例子示例了与现在的用于分布大文件的系统有关的局限性和问趙.因为这些问題,所以存在着对于使用更有效的方式来交付大的有效负栽的方法和设备的需求. The example above shows the limitations of the current system of large files for distribution and asked about Zhao. Because of these problems, so there is a need for the use of a more efficient way to deliver large effective method and apparatus for the negative planted.

发明内容 SUMMARY

本发明的一个实施例提供了用于遍及计算机网络来分布大文件和将这样的文件交付给最终用户系统的改进的机制.当本发明被实施时,它为多个用户提供了一种在不使网络资源负担过度的情况下获得对大的有效负栽文件的访问的方式.例如,如杲用户希望下栽一个大文件例如視頻文件,本发明的一个实施例提供了一种在不给网络造成紧张的情况下把那个視頻文件交付给请求的用户的方式.系统通过把大文件分成多个部分并把那些部分存储在遍及网络分布的位置(例如,节点)中来实现这一点.使用为大数据文件的智能管理提供的流程优化技术来分布遍及网络存储的各部分.这样,大数据文件的各部分被存储在使得把该部分交付给最终用户系统所花费的总的时间最小的位置.这些位置被本领域的普通技术人员称为网络的边缘. One embodiment of the present invention provides a mechanism for distributed throughout a computer network and improved large files such files delivered to end users of the system. When the present embodiment of the invention, which provides a plurality of users without gain access to a large payload file in the case where the plant network resources overtaxed manner. for example, the user wishes to Gao tilted downward and crashed as a large file such as a video file, an embodiment of the present invention to provide a network without user-friendly manner in the case of tension of the video file that is delivered to the requesting system by the large file into multiple portions and storing those portions distributed throughout a network location (e.g., nodes) to accomplish this. use of intelligent management of large data files provided by the process optimization techniques to distribute portions throughout the network storage. Thus, portions of large data files is stored in such storage to the portion delivered to the minimum total time of end-user systems takes place. these positions are those of ordinary skill in the art known as edge network.

具体实现本发明的各方面的在网络边缘上的每一个节点被配置为 Specific implementation of the invention each of the nodes on the edge of the network is configured to aspects of

当文件的各部分被真正地存储在遍及网络的其他节点上时看起来好象它具有本地存储的大文件.这在没有消耗系统资源的情况下大大地增 Large portions of a file when the file is actually stored on other nodes throughout the network looks as if it has stored locally. This greatly increased in the absence of consumption of system resources

加了每一个网络节点的虚拟存储能力.当最终用户系统发出对内容(例如,大的数据文件)的请求时,请求被发送到最近的节点,系统以在最小化带宽消耗的同时使得数据传送效率最大化的方式交付所请求的内容到该节点.最终的结果是每一个网络节点在不必本地地存储那些数据文件中的每一个的情况下访问到了大量的大数据文件, Adding the virtual storage capacity of each network node. When an end-user system issues a request for content (e.g., large data files), the request is sent to the nearest node, the system while minimizing bandwidth consumption such that the data transfer way to deliver content to maximize the efficiency of the request to the node. the end result is to visit each of the case without having those data files stored locally in each network node to a number of large data files,

在本发明的一个实施例中,系统被优化使得可以使用传送层网络袭盖来在现有网络(包括因特网和企业内部网)之间分布大有效负栽文件,以便把内容推进到网络的边缘.特別地,本发明的实施例提高了大有效负栽交付性能、可伸缩性、可靠性和可用性. In one embodiment of the present invention, the system is optimized so that the transport layer network may be used to cover the passage of a large payload distribution file planted between the existing networks (including the Internet and intranets), in order to advance to the edge network content in particular, embodiments of the present invention is to improve the delivery of a large payload plant performance, scalability, reliability, and availability.

如上所述,本发明的一个实施例把大有效负栽文件分成多个部分.这可以通过有选择地把大有效负栽文件分割成被复制和分配到位于网络边缘上的多个分布通信站(又名节点)的块,每一个分布通信站被配置为基于诸如用途、流行性等信息来确定多少内容被本地保存.内容提供者定义什么样的分布通信站作为分布通倌站是合格的,并且也可以定义其它的分布标准.网络中的分布通信站管理存储器和互相传送内容(例如大有效负栽文件的各部分)和其它的信息.大有效负栽文件的不同片段可以从不同的节点获取,但是,例如当用户请求通过一个应用服务器(例如,流型服务器)访问该大有效负栽文件时,虚拟文件控制系统产生整个文件现在就在所连接的节点上的错觉.但是,由于仅仅大有效负栽文件的被选择的部分在请求时可能实际地驻留在节点的存储器 As described above, the present embodiment of the invention the plant a large payload file into multiple portions. This may be copied and assigned by selectively dividing the large payload file into plant distributed to a plurality of communication stations on the network edge (also known as nodes) blocks, each distribution station is configured to communicate based on factors such purposes, epidemic and other information to determine how much the content is locally stored content provider defines what kind of distribution of the distribution through a communication station as a qualified station groom , and may also define other distribution criteria network distribution communication station management memory and various fragments transfer content (e.g., portions of large payload plant files) to each other, and other information. large payload plant files from different node obtains, however, for example, when a user requests via an application server (e.g., flow pattern server) to access the large payload plant file, a virtual file control system generates the entire file is now illusion on connected nodes However, since the only the selected portion of the large payload file when planted request may actually reside in the memory node ,分布通信站可以在该应用服务器服务于该用户时下栽该文件的没有驻留的部分.没有驻留的块可以并行地并且通常从最少拥堵的节点来下栽.整个过程对用户是透明的. , Not part of the resident file when tilted downward and crashed distribution of a communication station can serve the user in the application server may block does not reside generally parallel and tilted downward and crashed node from the least congested. The entire process is transparent to the user .

被请求文件的被请求部分被接收并且使用一个或多个被称为虛拟文件控制系统服务器的相关服务器实时地重新汇编.虚拟文件控制系统把重新汇编的文件提供给服务于客户机的应用服务器.虚拟文件控制系统可以被实现为可堆叠的文件系统,作为使用基本的网络文件系 Server is requested to the relevant part of the requested document is received and is referred to using one or more virtual file control system server reassembled in real time. The virtual file control system of the reassembled document to the service server to the client application. virtual file control system may be implemented as a stackable file system, as a network file system using the basic

统例如NFS或CIFS的代理文件服务器、存储区域网络(SAN)或直接访问的存储器,或者实现为这些方法的组合.无论哪一种实现被使用,虚拟文件控制系统从基本的文件系统获得该内容, System such as NFS or CIFS file server proxy, a storage area network (SAN) or direct memory access, or as a combination of these methods. Whichever implementation is used, the virtual file control system obtains the content from the basic file system ,

可伸缩内容交付网络通信站在地理上分散到网络的边缘以便最好地为位于边缘外的最终用户客户机系统服务.最终用户客户机对数据的请求自动地由最近的最不拥堵的通信站来服务.在本发明的一个或多个实施例中,可伸縮内容交付网络被综合到现有的在因特网的边缘的服务中来利用这些厫务(例如,在本发明的一些实施例中的应用服务器可以是在服务提供者的现有系统基础中运行的流型服务器). Scalable Content Delivery Network communication station geographically dispersed to the edge of the network in order to best serve the end user client system located on the outer edge of the service end-user client request for the data automatically by the nearest communication stations most uncongested to service this embodiment, the scalable content delivery network is integrated into an existing service to the edge of the Internet service using the Ao (e.g., in some embodiments of the present invention in one or more embodiments of the present invention. the application server can be a server running flow pattern in the existing system infrastructure service provider's).

在一个或多个实施例中,新的节点可以在没有服务中断的情况下被添加到网络中.随着新的节点的添加,它们从网络中的其它节点了解到它们应该有什么内容并且把所需的期望数重的内容从最近的和最少拥堵的节点下栽到它们的本地存储器上.这样, 一个节点可以被添加到网络中,并且在自预置后被启动并运行. In one or more embodiments, new nodes may be added without interruptions to the service network. With a new node is added, they learned from the other nodes in the network and what they should be put a desired weight of the desired number of contents tilted downward and crashed node from the least recent congestion and to their local memory. Thus, a node may be added to the network, and after the start and run from the preset.

在一个或多个实施例中,保留在每一个节点的大有效负栽文件的部分和总量依賴于可使用的存储器、内容的流行性、内容提供者的分布条件等.这样,大有效负载文件的最不可能被使用的块可以被删除(即,从本地存储器删除)来为其它高度期望的内容腾出空间.但是,尽管文件的最不可能被使用的块被删除,只要内容提供者希望内容保留在网络中,大有效负栽文件的全部内容可以被保留在可伸縮内容交付网络中的一个节点上. In one or more embodiments, each node remains in a large payload file portion and a plant depends on the total amount of memory may be used, the content popularity, content provider's distribution conditions, etc. Thus, large payload the most likely to be used blocks of a file can be deleted (i.e., deleted from local storage) to the other is highly desirable to make space. However, despite the block most likely to be using the file is deleted, as long as the content provider hope contents remain in the network, the entire contents of a large payload plant files can be kept on a node in a scalable content delivery network.

附图说明 BRIEF DESCRIPTION

闺l是内容交付的超高速緩存方法的一个示例. Gui l is an example of cache method of content delivery.

困2是连接到播放器的典型的流型服务器的一个示例.图3是使用镜象法把内容推进到网络逸缘的网络内容交付方案的一个示例. 2 is an example of a typical storm flow pattern server connected to the player. FIG. 3 is a mirror image method using a sample advance to the content delivery network content program Yi edge network.

困4是使用多点传送技术把信息推进到网络上的多个服务器的分布网络的一个示例. 4 is trapped using multicast technology to advance information to a plurality of exemplary distributed network of servers on the network.

困5是依振本发明的一个实施例的用于交付大有效负栽的可伸缩内容交付网络的一个示例. 5 is trapped by the vibration of one embodiment of the present invention, for example a negative large effective plant scalable content delivery network delivery.

困6是依据本发明的一个实施例的用于控制信息通信的节点的虛拟树结构的一个示例. 6 is an example of a trapped node information for controlling the communication of the virtual tree structure in one embodiment of the present invention basis.

困7是依据本发明的一个实施例的分布中心的简化布局. 7 is a simplified layout trapped distribution center in accordance with one embodiment of the present invention.

闺8是本发明中使用的线性和非线性文件结构的一个示例. Gui 8 is an example of linear and nonlinear file structure used in the present invention.

困9显示了依据本发明的一个实施例的把大有效负栽文件分解成块文件来存储的过程. 9 shows the difficulties of the present invention in accordance with one embodiment of the plant the large payload file into block files to decompose the stored procedure.

困io是依据本发明的一个实施例的被分解的大有效负栽文件的一 Io trapped present invention is based on a decomposed large payload embodiment of a plant file

个示例. One example.

困ll是显示从多个块文件重构一个大有效负栽的示意困. Ll is a reconstructed trapped a large payload trapped a schematic block file from a plurality of plant.

困12是依据本发明的一个实施例的属性位困和积累(rolled up)位闺的一个示例. 12 is an example of a trapped position based on the present invention Gui attribute bit trapped and accumulated (rolled up) of the embodiment.

困13是一个大有效负栽文件在本发明的网络中的分布的一个例示性实施例. 13 is trapped a large payload file planted a distributed embodiment of the present invention, in the network in an exemplary embodiment.

困14是可伸缩内容交付网络的另一个实施例的例示性例子. Illustrative examples of embodiment 14 is another embodiment of trapped scalable content delivery network in the embodiment.

困15是依据本发明的一个实施例的可伸缩内容交付网络通信站的 15 is trapped delivery network communication station based on the contents of a retractable embodiment of the present invention.

一个示例. An example.

困16提供了困14的可伸飨内容交付网鉻的一个替代示例. 16 stuck stuck 14 provides stretch satisfy a chromium content delivery network alternative example.

困17是依据本发明的一个实施例的分布服务器簇配置的一个示例. 17 is an example of a configuration trapped basis distribution server cluster to an embodiment of the present invention.

困18A-18C提供了依据本发明的实施例的应用服务器簇的三个例示性实施例. 18A-18C provides trapped application server cluster in accordance with embodiments of the present invention three exemplary embodiments.

闺19用于示例依据本发明的一个实施例的虚拟文件控制系统服务器的採作. Gui virtual file 19 according to an exemplary embodiment of the present invention, a control system for mining server.

困20是依据本发明的一个实施例的在可伸缩内容交付网络中的一 20 is trapped accordance with the present invention, an embodiment of the scalable content delivery network in a

个通信站的控制单元和数据的一个示例. One example of the control unit and data communication stations.

图21A到21G是依据本发明的实施例的通信站操作和数据流的例示性例子. 21A to 21G are illustrative examples based operations and data communication station of the present invention, embodiments of flow.

困22是依据本发明的一个实施例的在VFCS初始化过程期间执行的VFCS聚务器的操作的流程闺. 22 is a flowchart trapped Gui VFCS polyethylene based service is performed during the VFCS initialization process of the present invention, one embodiment of the operation.

困23显示了在运行期内执行的VFCS默务器操作. 23 shows the VFCS trapped default service operation performed at runtime.

困24是困20的内容管理和使用数振库的内容的示例. 24 is an example of sleepy sleepy 20 content management and content using a number of vibration library.

困25示例了在可伸缩内容交付网络中新的节点是如何通过了解和下栽它们应该具有的内容来自适应地初始化的, Examples of the 25 trapped in a scalable content delivery network through a new node is to understand how and content they should have tilted downward and crashed adaptively initialization,

具体实掩方式的详细梯述 Specific details described ladder cover embodiment

本发明的一个实施例提供了一个用于遍及计算机网络分布大文件(被称为大有效负我)和把这样的文件交付给最终用户系统的改进的机制.在下面的描述中,阐述了很多具体的细节来提供本发明的实施例的更详尽的描述.但是,对于本领域的普通技术人员显而易见的是,没有这些具体的细节本发明也可以被实施.在其它的例子中,众所周知的特征没有被详细描迷以便突出本发明. One embodiment of the present invention provides a computer network for distribution throughout the large files (referred to as a large payload I) and the delivery of such documents to an improved mechanism for end-user systems. In the following description, numerous specific details are many specific details to provide a more detailed description of the embodiments of the present invention. However, those of ordinary skill in the art to be apparent without these specific details may be implemented according to the present invention. in other instances, well-known features It has not been described in detail so as to protrude fan of the present invention.

当根据本发明的一个实施例实施本发明时,它为最终用户系统提供了一种访问大有效负栽文件而不会使由最终用户系统用来传送数据的网络超负荷的方式.在本发明的一个实施例中,系统通过把大有效负栽文件分割成多个部分并且把那些部分存储在遍及网络分布的位置(例如,节点)中.遍及网络存储的所迷部分使用为大有效负栽文件的智能化管理提供的流程优化技术来分布.这样,大有效负栽文件的各部分被存储在使得将所述部分交付给最终用户系统所花费的时间最小的位置中.这些位置使得与将该文件交付给最终用户系统相联系的等待时间最小化,在这里被称为网络的边缘. When, according to one embodiment of the present invention when the embodiment of the present invention, there is provided a system for end-user access to a large payload files without the plant data transfer mode network overload the system used by the end user. In the present invention, in one embodiment, the system is divided by the large payload file into multiple portions and planted those portions stored throughout the distributed network location (e.g., nodes). stored throughout the network using a large portion of the fans payload plant intelligent management file optimization techniques to provide a flow distribution. Thus, portions of the large payload file are stored in the plant such that the storage portion be delivered to the minimum time it takes the position of end-user systems in these positions such that the the document delivered to the end-user systems linked to minimize latency, referred to herein as the edge of the network.

实现本发明的各个方面的在网络边缘的每一个节点被配置为当文件的各个部分实际上被存储在位于遍及网络的其它节点中时看起来好象它具有本地存储的大有效负栽.这在不消耗系统资源的情况下大大地增加了每一个网络节点的虛拟存储容量.当最终用户系统发出对于内容(例如,大有效负栽)的一个请求时,请求被传递到最近的节点并且系统以在最小化带宽消耗的同时最大化数据传送效率的方式把所 Large implementations of the invention each of the edge nodes in the network are configured to various aspects of the various portions when the file is actually stored in other nodes located throughout the network appear as if it has a negative effective local storage plant. This without resource consumption greatly increases the virtual storage capacity of each network node. when an end-user system issues a request for content (e.g., large payload plant), the request is transmitted to the nearest node and the system manner while minimizing bandwidth consumption to maximize the efficiency of the data transmission

请求的内容交付给该节点.最终结果是每一个网络节点都访问了很多大数据文件而不必在本地存储那些数据文件中的每一个.这样,本发明的一个或多个实施例提供了不带有与现有技术的内容交付方法相联系的成本、可伸缩性、加栽均衡和其它问趙的将大有效负栽交付给网络的边缘的有效的方法和设备. Delivering the requested content to the node. The end result is that each network node visited many large data files without having to store those data files locally in each of the above, the present invention provides one or more embodiments without there are costs associated with the method of the prior art content delivery, scalability, plus planted balanced asked Zhao and other large payload will be delivered to the plant an effective method and apparatus for the edge of the network.

困s提供了依据本发明的一个实施例的用于交付大有效负栽的可伸缩内容交付网络(SCDN)的梘困.SCDN 5柳可以是一个诸如因特网的网络,它在概念上包括一个网络核心邻S(即主干)、离核心"近" 和"远"范闺内的中间网络段510和离核心"远"的网络段520A到520C (统称为520)."近"和"远"与距离有关并且想要指示到核心的相对路径等持时间(分別是短或长),这样的等待时间一般依赖于所经过的以到达形成网络核心的高速主干并且通过其传递大多数网络通信量的中间网络集线器(例如,交換机、路由器等)的数量.注意,每一个中间两络集线器可以在把所述通信量转发到下一个网络集线器之前执行一些有限的、增加了等待时间的处理. S trapped provide a basis for one embodiment of the present invention is the delivery of a large payload plant scalable content delivery network Liu Jian trapped .SCDN 5 (SCDN) may be a network such as the Internet that conceptually includes a network core o S (i.e. the backbone), from the core of the "near" and "far" intermediate network segments within the range Gui 510 and away from the core "far" network segments 520A to 520C (collectively 520). "near" and "far" For indicating the desired distance and the relative path to the core holding time and the like (short or long, respectively), which generally depends on the latency to reach the high-speed backbone network core is formed and through which the majority of network traffic passing through it number of intermediate hubs (e.g., switches, routers, etc.). Note that each intermediate hub may perform some network two limited, increasing the processing latency before forwarding the network traffic to the next hub.

困5显示了多个内容提供者客户机(CPC)系统53(J、多个最终用户客户机(EUC)系统550和一个或多个内容管理服务器(CMS) 570, 所有的这些都位于网络边缘501以外.通常,内容提供者客户机S30 可以被连接(或分配)到一个内容管理腹务器570,它依次被连接到其指定的分布中心540.内容提供者通过其CPC530上栽和/或管理SCDN500中的大有效负栽文件.EUC5邻提供对SCDN500中的文件的最终用户访问.例如,EUC550可以是在最终用户的本地计算机上运行的一个浏览器. 5 shows a plurality of trapped content providers client (CPC) system 53 (J, a plurality of end-user clients (EUC) system 550 and one or more Content Management Server (CMS) 570, all of which are located at the network edge other than 501. typically, the content provider client S30 can be connected (or assigned) to a content management service 570 belly, which in turn is connected to a distribution center 540. the content provider designated by CPC530 planted thereon and / or large management SCDN500 the payload file .EUC5 neighbor planted to provide end-user access to SCDN500 the file. for example, EUC550 may be a browser running on the end user's local computer.

网络边缘邻1一般地可以远离网络核心505.但是,核心和逸缘之间的距离(即珞径等待时间)不是相同的,并且对于给定的CPC或EUC 变化相当大.本发明的一个实施例在网络边缘放置了多个用于保持大有效负栽的分布中心(DC) 540A-54M,从而解决了等待时间问趙.来自内容提供者的大有效负栽内容从一个分布中心推进到位于网络边缘的另一个分布中心.寻求对大有效负栽的访问的最终用户由包含合乎要求的内容的最近的分布中心(通过一个应用朦务器)来服务.通过把内容邇过多个应用服务器560和位于边缘的分布中心540分布到最终用户(例如,EUC 550上),路径等待时间被最小化,这样,大有效负栽分布包括从内容提供者获得一个大有效负栽文件并且在地理上把这样的文件放置在位于边缘上的或尽可能接近网络边缘的分布中心上. Network edge o 1 may generally away from the core network 505. However, the distance between the core edge and Yi (i.e. latency Luo diameter) are not the same, and vary considerably for a given CPC or EUC. An embodiment of the present invention. EXAMPLE placed at the network edge for holding a plurality of large payload distribution center planted (DC) 540A-54M, so as to solve the latency Q. Zhao. from the content provider's large payload content advancing from one plant to a distribution center located another distribution center network edge. seek end-user access to a large payload planted by the nearest distribution center contains content desirable (through an application service is indistinct) to serve through the contents Wani had multiple application servers 560 distribution center and at the edge 540 of the distribution to the end user (e.g., the EUC 550), path latency is minimized, so that a large payload plant distribution comprises from the content provider to obtain a large payload plant files and geographically such files are placed in the edge-located on or near the edge of the network distribution center as possible.

例如,困5的SCDN500中的分布中心540A-S40I以闺6中示例的树的形式虚拟排列.这个虚拟树结构主要用于控制信息在可伸缩内容交付网络的节点之间的通信.数据下栽可以从具有期望的数据的网络中的任何节点、最好是最近的节点(在距离方面)执行.闺6的节点A 到I分别代表DC 540A到5401.节点以逻辑顺序排列.例如,假设节点B代表欧洲-英闺,則在欧洲的逻辑子节点可能是欧洲-法国(例如, 节点D)和欧洲-德闺(例如,节点E),欧洲-法国的子节点可能是欧洲-意大利(例如,节点H).在这个例子中,树的左倒代表欧洲,右倒可以代表亚洲.节点A是根节点,例如可以代表中心控制站.在一个或多个实施例中,树中的每一个节点具有代表节点的名字的唯一的属性集.对于一个节点的属性集被存储在节点中并且可以用任何常规的数据结构表示.例如,属性集可以被表示为可变的 For example, the distribution center 5 SCDN500 trapped 540A-S40I in the form of a tree of virtual Gui exemplary arrangement 6 This virtual tree structure information is used to control the scalable content delivery network communication between nodes of data in a tilted downward and crashed from any node in the network having the desired data is preferably performed nearest node (in terms of distance). node a to I Gui 6 represent DC 540A to node 5401. in a logical order. For example, assume that node B represents Europe - British women's apartment, then the logical sub-node in Europe could be Europe - France (for example, node D) and Europe - Germany boudoir (for example, node E), Europe - France child nodes may be Europe - Italy (for example, , node H). in this example, the tree inverted behalf of the European left, and right down may represent Asia. node a is the root node, for example, may represent a central control station. in one or more embodiments, each of the tree node unique set of attributes having the name of the primary node. attribute set for a node is stored in the node and can be represented by any of conventional data structures. for example, the attribute set can be represented as a variable 图(位困是一个目标、例如一个数字的二进制表示).每一个节点也包含每一个节点的子节点、孙节点、重孙节点等(即,从作为一个根节点的那个节点发源的所有节点一直系子代)的属性集的表示.这个表示被称为"积累属性集"并且常规的数据结构都可以用于它.这样, 一个节点的积累属性是它的子节点的积累属性的表示.例如,可以使用一个"积累位困",它是该节点的所有子节点的积累属性位困的组合. 一个"巻积位困"可以被定义为节点的子节点的积累属性的"二进制OR"(也称为逐位OR).困12是根据本发明的一个实施例的属性位困和积累位闺的一个示例.出于例示目的,位困1200、 1210、 1220、 1230、 1240 和1250使用16位,但因为位困是可变的,它们可以根据需要变化以识别每一个节点并且提供其它必要的信息. FIG. (A target bit is stuck, for example, a digital binary representation.) Each node also contains a node for each child node, and a grandchild node, great-grandchildren nodes, etc. (i.e., all nodes from that node as a root node of origin progeny been represented) set of attributes. this representation is called "accumulated property set" and conventional data structure may be used for it. Thus, the attribute of a node is accumulated accumulation its child nodes represent attributes. For example, a "accumulated bit storm", it is the accumulation of attribute bits for all child nodes of the node trapped combination a "convolving bit trapped" may be defined as the accumulated attribute child node of "binary OR" (also known as bitwise OR). trapped attribute bit 12 is one example of the Inner trapped and accumulating bits according to one embodiment of the present invention. for illustrative purposes, stuck bits 1200, 1210, 1220, 1230, 1240 and 1250 used 16, but because the trapped bit is variable, which may be to identify each node and provide other varied as desired necessary information.

表示困6的节点B的属性集的位困1200具有作为它的标识的被设置为1的位l、 4和13和被设置为O的所有其它位.因为节点B是A 的一个子节点,所以可以设置位l,例如,位4可以被设置为表示欧洲, 位13被设置为表示英国.表示作为B的一个子节点的困6的节点D的属性集的位图1210具有被设置为1的位1、 4和14和被设置为0的所有其它位.例如,位14可以表示法国.表示也是B的一个子节点的图6的节点E的属性集的位图1220具有被设置为1的位l、 4和15和被设置为O的所有其它位.例如,位15可以表示德国.表示作为D的一个子节点的困6的节点H的属性集的位困1230具有被设置为1的位l、 4和16和被设置为0的所有其它位.例如,位16可以表示意大利.如前面所讨论的,节点D的积累位田(例如,1240)将作为节点H的属性位困(因为H没有任何子节点),节点B的积累位困(例如,12邻) 是位困1210 Bits representing the storm 1200 attribute set trapped Node B 6 having as its identifier is set to bit l 1 and 4 and 13 and is set to O all other bits. Since the node B is a child node of A, it is possible to set the position L, for example, bit 4 may be set to represent Europe, bit 13 is set to represent England. bitmap representing the attribute set of node D trapped 6 as B is a child node 1210 has been set to 1 bits 1, 4 and 14 and are provided to all the other bits to 0. For example, bit 14 may represent France. bitmap representing the attribute set of FIG child node is node B E 6 1220 has been set to 1 bit l, 4, and 15 and are provided to all the other bits O For example, bit 15 may represent Germany, bits representing trapped attribute set as D is a child node trapped node H 6, 1230 has been set to 1 bit l, 4, and 16 and are provided to all the other bits to 0. For example, bit 16 may represent Italy. as previously discussed, the node D accumulation of bit fields (e.g., 1240) as attribute bits trapped node H ( because H not have any child node), the node B stuck accumulated bit (e.g., 12 o) are trapped position 1210 1220和1230的二进制OR. 二进制OR的结果是在位困1210、 1220和1M0中设置的所有的位也被设置在积累位困1250 (即, 位l、 4、 14、 15和16)中. 1220 and 1230 binary OR. OR result of the binary bit trapped 1210, 1220 and all the bits are also provided 1M0 accumulation bit provided sleepy 1250 (i.e., bits l, 4, 14, 15 and 16).

内容管理服务器S70可以被连接到树上的任何一个节点.这样,尽管一个内容管理服务器和一个分布中心可能未被并置在同一处,内容管理服务器为内容提供者提供了一个栽体来把大文件(例如,视频) 上载到分布中心.在一个实施例中,内容管理服务器是一台处理内容提供者的大有效负栽文件以供在网络中分布的计算机.在另一个实施例中,内容管理服务器例如可以是允许把内容上栽到网络的工具的子集(例如,与机器无关的目标);这样,这些工具可以从一个服务器运送到用于网络中的大有效负栽文件的处理和分布的内容提供者客户机的计算机.在内容提供者把大有效负栽文件加栽到内容管理服务器中后,CMS可以处理该文件并且把它转发到分布中心. S70 content management server can be connected to any node in the tree. Thus, although a content management server and a distribution center may not be collocated in the same place, the content provider to the content management server provides a plant body to the big files (e.g., video) uploaded to the distribution center. in one embodiment, the content management server is processing a large content provider planted payload file for distribution in the network computer. in another embodiment, the content for example, the management server may be a subset allows plummeted (e.g., a machine-independent target) tools network content; Thus, these tools can be transported from one server to the processing plant the large payload file in the network and for after the distribution of the content provided by the client's computer. the content provider to large payload files plus plant plummeted content management server, CMS can handle the file and forwards it to the distribution center.

根据本发明的一个实施例,在困7中示例了分布中心的一个简化布局。 According to one embodiment of the present invention, an example of a simplified layout in the center of the distribution of trapped 7. 分布中心700包括控制羊元701、 一个或多个虚拟文件控制系统702、 一个或多个分布服务器703和多个存储设备(例如,711-713). 控制单元701是分布中心的网络管理者,它的功能将在后面的部分中进一步讨论.不属于分布中心700的一部分的应用服务器721-724 (例如,流型服务器、FTP服务器和媒体播放器)在本示例中被显示连接到虚拟文件控制系统702以提供对于最终用户客户机如何访问存储在SCDN中的大有效负栽文件的可见性.分布服务器700的部件可能没有被布置在相同的节点中.例如,VFCS 702可以与应用服务器(例如, 721-"4)在一起,控制单元(例如,CU701)可能位于别处,例如与VFCS702在一起.这样,对于分布服务器700的所有部件不需要共同位于一个SCDN节点上. Distribution center 700 comprises a control element 701 sheep, one or more virtual file control system 702, one or more distribution server 703 and a plurality of storage devices (e.g., 711-713). The control unit 701 is the center of the distribution network manager, its function will be discussed further in the following section. not part of distribution center application server 700 721-724 (e.g., flow pattern servers, FTP servers, and media players) are shown coupled to a virtual file control in the present example system 702 in order to provide on how end-user clients access large stored in the SCDN in payload plant file visibility. distribution server component 700 may not be disposed on the same node. for example, VFCS 702 may be an application server (e.g., , 721- "4) together, the control unit (e.g., CU701) may be located elsewhere, for example VFCS702 together. Thus, for distribution to all components of the server 700 need not be co-located on a SCDN node.

内容提供者使用在一个内容提供者客户机系统上运行的内容出版和管理工具把大有效负栽文件上载到一个单个的内容管理服务器.在接收到该文件后,CMS处理该文件并且如果需要的话把它分成磁道文件(也称为线性文件). 一个线性文件包括保持与文件的实质(即, 实质性内容)有关的類序.例如,如果线牲文件包含一部电影,文件的开头将包括电影的开头部分.类似地,电影的中间和结尾部分将位于线性文件的中间和结尾.例如,线性文件合乎要求是因为它更容易使用线性叠加来重新汇编. 一些媒体文件不是线性的,也就是,它们包含多个磁道,使得例如电影的笫一部分没有存储在文件的开始.在 Content providers use on a content provider client system content publishing and management tools to run the large payload planted upload files to a single content management server. After receiving the document, CMS to process the file and, if necessary it is divided into track files (also referred to as linear file). a linear file comprises a file holding the substance (i.e., substantive content) related to the class sequence. For example, if the file contains a movie offerings lines, including the beginning of the file the beginning of the film. Similarly, middle and end of the movie will be in the middle and the end of the linear file. for example, linear file desirable because it is easier to recompile using a linear superposition. Some media files are not linear, that is, , they comprise a plurality of tracks, for example, such that a portion of Zi stored in the file does not start the movie in

把该文件拆分成线性(即,磁道)文件后,CMS把该文件传送到它连接的分布服务器.分布服务器进一步把磁道文件拆分成块文件,成为合乎存储器要求的.例如,接着块文件可以存储在本地存储单元711-713中.接着使用文件分布协议(例如,FDP)命令来把该文件或其选定的部分分布(即,复制)到可伸缩内容交付网络中的其它分布服务器节点.对于最初的复制,整个块文件不需要被存储在所有节点中, To split the file into a linear (i.e., track) the files, the CMS transfers the file to the distribution server it is connected. Distribution server further tracks to split the file into block files, memory requirements become desirable. For example, block file then It may be stored then use the file distribution protocol in the local storage unit 711-713 (e.g., of FDP) command to the selected portion of the document or the distribution (i.e., copied) into the scalable content delivery to other distribution server nodes in the network for initial replication, the entire block files need not be stored in all nodes,

但是一个原版拷贝可以完整地保留在一个节点(典型地是源节点)中. FDP包括便于在SCDN内的文件传送和操作的命令.块的大小影响内 However a master copy may be intact in one node (typically the source node) in. Of FDP comprises facilitating file transfer operations and commands within the SCDN. Affect the size of the block

容分布和内容交付的性能并且将在这个文本的后面讨论. Content distribution and content delivery performance and will be discussed later in this text.

虛拟文件控制系统(VFCS) 702能从块文件拼回到一起成为原始(大有效负栽)文件.如同后面将解释的,大有效负载文件的所有块不需要存储在一个分布中心上,但是整个文件在SCDN中可用。 Virtual File Control System (VFCS) 702 fight back together from the block files into the original (large payload plant) files. As will be explained later, all the blocks necessary to store large payload files on a distribution center, but the entire file is available SCDN in. 当一个最终用户连接到应用服务器721 (例如流型服务器)时,VFCS产生 When an end user connects to application server 721 (e.g. streaming server type), VFCS generated

一个整个文件在那个节点可以得到的虚拟表现.例如,假设仅仅两个千兆字节的文件的50%被存储在存储器711-713中,VFCS使流型服务器721认为整个两千兆字节在该位置是可得到.这样,流型服务器"l 可以开始播放这个文件.随着文件被播放,VFCS与DS通信来从网络中的其它节点定位和检索该文件的剩余部分. An entire file in that node can be a virtual performance. For example, only 50% assuming two gigabyte file is stored in memory 711-713, VFCS makes streaming server 721 think that the entire two-type gigabytes this position is available. Thus, the flow pattern server "l may start playing the file. as the file is played, DS v FCS to communicate with and retrieve the remainder of the file from other nodes in the network positioning.

分解大文件 Break down large files

大有效负栽文件在若干步樣中被分成多个块,确切的过程依賴于它是否是线性文件或非线性文件.例如,使用一个电影文件,如果这个电影的第一个10%近似地位于文件的笫一个10%中、下一个10%位于文件的下一个10%中,等等,則该文件是线性的.相反,在电影的第一个10%位于不是文件开头的其它地方的电影文件被认为是非线性文件.线性和非线性文件结构的例子示例在困8中.数据格式800可以表示例如是线性的mpeg格式,因为它从开头开始在一单个磁道中包含在整个文件中被多硌复合在一起的音頻/枧頻数振,注意,各种格式的每一个子部分表示一个磁道,因此格式810»830中的每一个包含多个磁道.如困所示,格式810是非线性的,因为它在文件的第一磁道中包含首部信息,接着在下一个磁道中跟着有元信息,然后在笫三磁道中包含着視頻信息,然后在第四磁道中 Large payload file is divided into a plant sample in several steps in a plurality of blocks, the exact process depending on whether it is a linear file or a non-linear file. For example, using a movie file, if the first 10% of the movie is located approximately 10% of the undertaking of a file, the next 10% at 10% is located in a file, etc., the file is linear. in contrast, other parts of 10% in the first film at the beginning of the file is not a movie file is considered to be non-linear file. Illustrative examples of linear and non-linear file structure data format 800 may be represented in a linear trapped 8 mpeg format, for example, from the beginning because it is contained in a single track are multiplexed in the whole file Ge together composite audio / vibration frequency soap, note that each of the subsections of various formats represent a track format thus each comprise a plurality of tracks 810 »830. as shown trapped, format 810 is non-linear because it contains a first track in the file header information, and then followed by the next track meta information contains video information and undertaking of three tracks, and a fourth track 含着元信息,在第五磁道中包含着第一音頻信道,在第六磁道中包含着笫二音頻信道,然后在最后包含着一些控制信息.这样,格式化为格式810的电影的开头将不位于文件的开头.格式820和830是其它可能的非线性媒体数振格式的表示.例如,格式820可以使数据被格式化为使得该文件在开头包含首郜信息,然后是诸如mpeg的格式的一些56K编码,后面跟着128K 的编码信息.其它媒体格式830可以包含首部信息,后面跟着索引信息,后面跟着視頻,最后是音頻信息.为了与在这个说明书的后面要讨论的复制算法兼容,所有的这些和其它非线性文件需要首先被转換成线性文件. Mouth meta-information included in the fifth track with a first audio channel, Zi contains two audio channels in the sixth track, and the last contains some control information. Thus, the formatted film 810 at the beginning of the format is not located at the beginning of the file. formats 820 and 830 are other possible non-linear media format indicates the number of vibration. For example, the data format 820 can be formatted such that the file contains header at the beginning Gao information, such as an mpeg format and some 56K encoding, followed by coding information of 128K. other media format 830 may contain header information, followed by index information, followed by video, and finally audio information for compatibility with the replication algorithm in the back of this specification to be discussed, all these and other non-linear files need to first be converted to a linear file.

闺9显示了把大有效负栽文件分解成块文件以供存储的过程.在内容提供者把文件上栽到内容管理厥务器(CMS)之后,CMS判定文件是线性的还是非线性的.如果该文件是线性的(例如,块9S0),例如一个mpeg电影,CMS把数据发送到在块930上的DS进行分块处理. 然而,如果文件是非线性的(例如,块900),则CMS在块910执行去多路复合过程以产生线性磁道文件920.去多路复合过程涉及将非线性(即,多个磁道)文件分解成分别包含单个磁道的文件.例如,采用如困10所示的媒体数据作为示例,大有效负栽文件1000包含笫一磁道中的首部、第二磁道中的視频、笫三磁道中的笫一音频信道、第四磁道中的笫二音頻信道以及最后的第五磁道中的控制信息.内容管理服务器将大有效负栽文件1000拆分成五个线性磁道文件1010,使得一个文件包含首部,笫二个文件包含視頻数据,第三 Gui 9 shows the decomposed large payload file into block files plant for the stored procedure. After plummeted to the content provider device content manager Jue Service (CMS) on the document, CMS is determined that the file is linear or non-linear. If the file is linear (e.g., block 9S0), for example a mpeg movie, CMS sends the data to the DS at the block 930 into blocks. However, if the file is non-linear (e.g., block 900), then the CMS in block 910 to execute the process to generate a multiplexed linear track file to the multiplexer 920. the process involves a non-linear (i.e., multiple track) files are broken down into files containing single tracks. For example, with 10 as the storm media data as an example, a large payload plant file 1000 contains header Zi track, the second track of the video, Zi three tracks Zi audio channels, fourth tracks undertaking of two audio channels and the final the control information in the fifth track. the content management server planted large payload file 1000 into five linear track files split 1010, a file that contains a header, Zi two file including video data, third 个文件包含第一音频信道,以此类推. File contains the first audio channel, and so on.

再参考困9,线性磁道文件920或线性大有效负栽文件9S0 (也是一个线私磁道文件)由CMS在网络上发送给它所连接的一个DS.文件可以根据下面讨论的文件分布协议(FDP)来发送,将来自CMS的文件输入到一个基于DS的分块过程930,基于DS的分块过程930产生块文件940.块文件940随后被存储在DS的本地存储器中.在处理之后,内容可以由网络中的其他分布服务器下栽. 一般地,在网络上传送的文件的大小与存储在DS的本地存储系统中的块文件之间不需要有直接的关系. Referring again trapped 9, linear or linear track file 920 planted large payload file 9S0 (track file is a private line) transmitted by the CMS over the network to a DS that it is connected. The file can be a file distribution protocol discussed below (FDP ) to send the file input from the CMS to a DS-based process of block 930, the file generating block 930 based on the DS block 940. process block 940 then the file is in the local memory of DS. after the processing, the content storage It can be tilted downward and crashed by other distribution servers in the network. in general, the size of the transferred file is stored on networks without a direct relationship between the local storage file system blocks in the DS.

分块过程930将磁道文件拆分成更小的可管理的单位,如困10的块1020中所示.分块过程产生多个块文件H、 V"、 A!.w.2、 Am2和C(统称为困10中的1020).块文件可以包含数据重叠或偏移(例如, 移位).例如,块文件V4可以包含首部磁道的一些部分,等等.在本发明的一个或多个实施例中对块文件的唯一的要求是每个磁道的开头被包含在为该磁道创建的第一个块文件中,例如,音頻信道1的开头被包含在A"中,音频信遣2的开头被包含在AL2中,等等.其他实施例可以在不首先经历去多路复合过程(例如,块910)的情况下仅仅将大有效负栽文件(即,非线性)直接拆分成块文件,从而每个块文件可以包含重叠的磁道.将大有效负栽文件拆分成块使得能够将块文件分布到不同的存储设备中,并在不影响系统性能的情况下在需要时添加新的存储设备.因此,例如,更多的存储设 Splitting process block 930 to track files into smaller manageable units, as shown in block 10 trapped. Process block 1020 generates a plurality of blocks in files H, V ", A! .W.2, Am2, and C (collectively, the trapped 101,020). block files may contain data overlaps or offsets (e.g., shift). For example, block file may contain some portion V4 header portion of the track, etc. in the present invention, or a in one embodiment the only requirement for the block files embodiment is the beginning of each track is contained in the first block file created for that track, for example, at the beginning of the audio channel 1 is contained in a ", the audio signal sent 2 AL2 is included in the beginning, and so on. other embodiments may not initially subjected to multiplexing to process (e.g. block 910) only in the case of a large payload file plant (i.e., nonlinear) direct Optical resolution block files, so that each block file may contain overlapping tracks. planted split a large payload file into blocks makes it possible to block the distribution file to a different storage devices, and added when needed without affecting system performance the new storage device. Thus, for example, more storage device 可以被添加到分布中心(困7),而不需要象现有技术中一样将文件移来移去或重新配置其他节点.例如,不同块可以位于SCDN的不同节点,因此位于不同的存储设备上.较小的块文件使得能够在同一时间支持多个应用服务器(例如,流型服务器),从而增大访问带宽.例如, 一个大有效负栽文件的多个块文件可以并行下栽.在未首先将整个文件下栽到流型服务器的情况下由用户进行的快进和快倒也是可能的. It may be added to the distribution center (trapped 7), without requiring the same as the prior art will be moved to remove a file or reconfigure other nodes. For example, different blocks may be located in different nodes of the SCDN, and therefore on different storage devices the smaller block files makes it possible to support multiple application servers (e.g., server-flow type) at the same time, thereby increasing access bandwidth. For example, a large payload file plant plurality of parallel block file tilted downward and crashed. without first, the entire case file to the flow pattern tilted downward and crashed server by the user fast-forward and rewind are possible.

从块文件重构大有效负栽文件 Reconstruction of large payload file from the plant block file

困ll是显示由VFCS从多个块文件重构一个大有效负栽文件的过程的示意困.块文件1100被输入汇编过程1110.如前面一个部分所讨论的,分块的逆过程被称为"汇编".虚拟文件控制系统(VFCS)使用汇编过程lllO来将多个块文件变换成线性磁道文件.如果原始大有效负栽文件是线性的,汇编过程1110只产生一个线性磁道文件(例如, 线性大有效负栽文件1150).然而,在原始大有效负栽文件是非线性的时,汇编过程1110产生多个线性磁道文件1120.线性磁道文件由适当的块文件的线性组合来产生.例如,困10的視頻磁道文件通过对块 Ll is a reconstructed trapped a large negative effective process plant file a schematic block file from a plurality of trapped by the VFCS. Block files 1100 are input 1110. The assembly process, the reverse process as a block portion is referred to as previously discussed "assembly." virtual file control system (v FCS) lllO assembly process using a plurality of blocks to transform files into linear track files. If the original large payload file is linear plant, the assembly process 1110 generates one linear track file (e.g., planting linear large payload file 1150). However, the large payload file in the original plant is non-linear, the assembly process to produce a plurality of linear track files 1110 1120. the linear track file is generated by a linear combination of the appropriate block files. for example, trapped by the video track file of the block 10

文件Vh V2、 V3和V4进行线性组合(即,求和)来重新产生.线性 File Vh V2, V3 and V4 linearly combining (i.e., summing) to reproduce Linear

磁道文件1120可以进一步在多硌复合过程1130中组合,以产生非线性大有效负栽文件1140.多路复合过穉仅仅对磁道文件进行重新汇编以产生原始非线性大有效负栽文件. Track files 1120 may further be combined in a multi-Ge recombination process 1130 to produce a non-linear large payload file plant through multiplexer 1140. Zhi only the track files to generate the original non-linear reassembled large payload file plant.

文件分布协议(FDP) File distribution protocol (FDP)

FDP协议定义了传送、存储和操纵内容提供者文件和存镛在网络中的文件元数据所需的文件管理基元.这种基元包括上栽、分布、交付、修改和刪除文件的命令.FDP命令产生在网络中的适当的服务器之间传送的一个或多个包.对于本领域普通技术人员来说显然的是, 这里所描迷的命令名字和协议实现是为方便而使用的,只要产生网络内有效的和可靠的丈件传送,其他命令或协议可以被添加、刪减或替代. FDP protocol defines the transfer, store, and manipulate content provider files and file metadata stored Yong required data in the network document management primitives. Such primitives include the plant, distribution, delivery, modify, and delete file command. FDP command is generated in a transmission network between one or more appropriate server packages for those of ordinary skill in the art it will be apparent that the fans described herein command names and protocol are used for convenience, as long as produce effective and reliable husband transported within a network element, other commands or protocols may be added, deletion or substitution.

"放置(Put)":内容提供者使用在内容提供者客户机系统上运行内容管理应用程序来将一个文件(内容)和文件元数据(与在网络 "Place (Put)": content providers use to run on the content in the content provider client system management applications to a file (content) and file metadata (and in the network

中被存储、传送和操纵的文件的管理有关的数据)上栽到内容管理服务器(CMS)上.CMS将文件分成线性磁道文件,然后向一个最终在网络中分布该内容的DS发出一个"放置"命令.在一个实施例中, CMS连接到在一个SCDN节点的DS. CMS为每一个磁道文件向DS 发送一个"放置"命令.事实上,"放置"命令是一个"推进"动作, 将一个磁道从一个CMS推进到一个DS."放置"命令可以包括四个包,例如:"放置"、"放置_响应"、"放置—组块"和"放置j答". "放置"包告诉接收DS准备好接收一个磁遣文件."放置一响应"包是一个由DS发出的包,向CMS指示该DS是否需要接收该磁道文件, 如果需要,从哪里开始传输.在一个磁道文件的一部分已经被传送之后通信对话期被打断并且CMS需要重新传送文件的剩余部分时的愔况下,这个包可能有用. 一旦DS告知CMS从哪里开始传送一个磁道文件,CMS可以与实际 Are planted to store data related to the management, transfer and manipulation of files) on the content management server (CMS) .CMS the file into linear track files and then issues a "to a final DS of the content distribution network is placed in the "command. in one embodiment, the CMS is connected to a DS in the SCDN node. the CMS for each track file to the DS sends a" place "command. in fact, the" place "command is a" promoting "operation, a a track from a CMS to advance DS "place" command may include four packets, for example:.. "place", "place _ response", "place - chunk" and "place j a", "place" packet tells the receiving DS is ready to receive a magnetic removal file. "placed a response" packet is a packet issued by the DS, the CMS indicating whether or not the DS needs to receive the track file, if necessary, start transmission where in a portion of a track file has been after the transfer of the communication session is interrupted and CMS need to re-transmit under conditions when the yin rest of the file, this package may be useful. Once DS inform CMS start transferring files from one track where, CMS can actually 的磁道文件一起发出一个"放置_组块"包. DS在接收到整个磁道文件时可以以一个"放置_应答"包作出响应, 以表明成功的传输.在接收到磁道文件之后,DS将线性磁道文件分成块文件,将块文件存储在本地存储器中,并更新文件元数据以反映该磁道、块和位置信息. Issuing a "Place _ chunks" package track files together. The DS when receiving the entire track file may be a "place _ response" packet in response to indicate the successfully transmitted. After receiving the track file, DS will linear track files into block files, in a local storage, and updates the file metadata to reflect that the block file storage track, block, and location information.

"分布(Distribute)":在所有磁道都已经被推进到DS之后, CMS可以发出引导DS将文件分布给网络中的其他节点的"分布"包. 例如,CMS可以为每个磁道文件发出一个"分布"包,每个包都包含内容提供者的分布标准.分布标准例如可以指定网络中的哪些节点应该具有该内容."分布"命令可以包括两个包,例如:"分布"和"分布—应答".DS可以通过向CMS发出一个"分布J答"包来对接收 "Distribution (Distribute)": After all of the tracks have been advanced to DS, CMS may issue a boot file to be distributed DS "profile" of other nodes in the network package, for example, CMS may issue a file for each track. " profile "packets, each packet contains a standard distribution of the content provider's distribution criteria, for example, may specify which nodes in the network should have the content." profile "command may include two packets, for example:" profile "and" distribution - answer ".DS can be issued to the CMS by a" J distribution answer "to the received packet

到"分布"命令和磁道文件作出肯定应答. To the "distribution" command affirmative answer file and track.

"复制(R印licate)":响应于"分布"命令,DS可以向其邻节 "Copy (R printed licate)": in response to the "profile" command, DS its neighbor section can

点发出"复制"包.每个满足由内容提供者规定的分布标准的邻节点可以向分布路径中的一个或多个DS发出一个命令(诸如下面描迷的 Point issuing "replicate" packets. Each satisfied by the content provider a predetermined distribution criterion neighbor may issue a command to the distribution path with one or more DS (as described below fans

"获取"包),以便将文件的一部分拉入其本地存储器."复制"包从已经推进了磁道文件的DS开始."复制"包用作对DS的关于它可能需要将来自任何一个发送的DS的特定块文件拉入(即,复制)到其本地存储器的通知.接收的DS可以通过发出一个"复制—应答"包来对该通知作出肯定应答,然后,它在准备好时承担从发送的DS拉入块文件的责任. 一个DS通过向它们发出"复制"包来进一步通知其相邻节点以确定它们是否应该拉入部分或整个文件.如杲积累属性匹配内容分布标准, 一个DS可以向其子代节点发出一个复制请求. "Get" bag), so that the portion of the file to pull into its local memory. "Replicate" packet starts from an already advanced track file DS. "Copy" package used for the DS on it may be necessary from any DS a transmitted the particular block file pull (i.e., replicate) to notify its local memory may be received by a DS "copy - response" by issuing. acknowledgment packet to make this notification, and then, when it is ready to take in the transmitted DS pull liability blocks of the file. DS by issuing a "copy" packets to them to further notify its neighboring nodes to determine if they should pull part or the entire file. the attribute matches Gao accumulation distribution criteria, may provide a DS descendant node issues a copy request.

"获取(Get)": —个需要从另一个DS拉入文件的DS例如可以发出一个"获取"命令."获取"命令可以包括四种类型的包:"获取"、"获取_响应"、"获取_组块"和"获取_应答".例如,"获取"包可以用于启动一个拉入,"获取_响应"包可以用于报告通信站的状态和在需要时传送文件元数据."获取一组块"包可以用于传送文件数据,"获取一应答"包可以用于对"获取"序列的结束作出肯定应答并报告状态.DS可以基于下列因素来决定要拉入的文件的大小:(1) 其存储器可用性;(2)网络闺中通信站的位置;(3)内容的流行性; "Get (Get)": - a need to pull files from another DS DS, for example, may issue a "get" command "get" command may include four types of packets: "get", "_ Get Response". "Get chunk _" and "_ Get response." for example, "get" packet may be used to initiate a pull, "_ Get response" packet may be used to report the status of communication stations transmitting and file metadata as needed. "Getting a chunk" of data packets can be used to transfer files, "get a response" package can be used to "access" to the end of the sequence and acknowledgment status report .DS can decide to pull the file based on the following factors size: (1) the availability of its memory; position (2) apartment in a network communication station; epidemic (3) content;

(4)文件的可截取或不可截取特性;以及(5)带宽容限.DS可以响应于"复制"请求和"搜索_回答"请求发出"获取"命令序列. Intercept (4) file or a non-intercepting characteristic; and (5) the bandwidth tolerance .DS response to the "copy" request and "Search _ answer" Request sending a "Get" command sequence.

"准备(Pr印are)":"准备"命令可以包括两个包,例如:"准备"和"准备—应答》 .节点的VFCS可以向一个DS发出一个"准备》 包以便为一个应用服务器拉入文件的未驻留部分.DS可以使用"准备_应答"包来应答它已经接收到"准备"包以及它将尽可能快地执行"准备". "Preparation (Pr are printed)": "prepare" command may include two packets, for example: "prepare" and "preparing - Answer" v FCS node may issue a "prepare" packet to a DS to pull to an application server. non-resident portion of the document may be used .DS "_ prepare response" packet it has received a reply to "prepare" packet and will execute as fast as possible "ready."

"搜索(Search)":当DS可以处理"准备"请求时,它可以发出一个"搜索"命令来定位一个文件的缺少部分."搜索"命令可以包括三个包,例如:"搜索"、"搜索—应答"和"搜索—回答".一个服务于"准备"命令的DS发出一个"搜索"包来启动在其邻节点中的对文件的未驻留部分的搜索.每个邻节点可以发出一个指示它已经接收到"搜索"请求的"搜索J答"包."搜索J答"包不是DS具有所请求文件的部分的肯定应答.具有所请求文件的一部分的节点可以发出一个"搜索_回答"包."搜索一回答"包可以包括诸如驻留在通信站中的所搜索文件的部分、通信站的网络条件以及通信站的DS簇的负栽之类的信息.启动DS簇中的一个DS接收"搜索—回答"包, 并且可以基于"搜索一回答"包中的信息选择合适的远程DS节点来下栽文件的缺少部分.启动DS簇中的一个DS例如可以向一个或多个通信站(即,选定的SCDN "Search (Search)": When the DS can process "prepare" request, it may issue a "search" command to locate the missing portions of a file, "search" command may include three packets, for example:. "Search," " Search - answer "and" Search - answered. "a service to" prepare "command DS issued a" search "package to start its neighbors in search of the file is not part of the resident can be issued for each neighbor. indicating that it has received the "search" request "Search J A." package. "Search J a" packet is not the DS having an acknowledgment portion of the requested file. node has the requested part of the file may issue a "search _ answer "package." search for a answer "packet may include information such as network conditions and DS cluster communication station searched files reside in a portion of a communication station, the communication station or the like planted negative start DS cluster DS receiving a "Search - answered." package, and may be based on the "search for a answer" select appropriate remote DS nodes package file missing portions tilted downward and crashed to a DS start DS cluster may be, for example, to one or more of communication stations (i.e., selected SCDN 点)发出"获取"命令来下栽缺少的内容. Point) issued a "get" command tilted downward and crashed to the lack of content.

"移去(remove)":"移去"命令可以包括两个包,例如"移去"和"移去一应答".节点控制单元可以向DS发出一个"移去"命令来移去某些块.后面将描迷的修剪过程使用该"移去"命令."移去"包是对DS的某些块必须被移去的通知.DS随后可以发出一个"移 "Remove (remove)": "remove" command may include two packets such as "remove" and "Remove a response" node control unit may issue a "remove" command to the DS to remove some of the. blocks will be described later pruning process using the fan "remove" command. "remove" packet is a DS certain blocks have to be removed can then issue a notification .DS "move

去一应答"包来应答:它将在准备好时最终移去所指示的块. A response to "package to answer: it will eventually remove the block indicated when ready.

"清除(clean)":"清除"命令可以包括两个包,"清除"和"清除_应答",CMS可以发出一个"清除"或类似的包以通知位于同一节点的一个DS它需要移去一个特定文件.DS发出一个"清除一应答"或类似的包来应答:文件将在准备好时最终被移去.按照在"复制"命令期间使用的路径(在该文件的分布标准中可得到的),DS向它的请求从SCDN中的所有通信站«除文件及其相关文件元数据的相邻节点发出一个"清除"或等效命令. "Clear (clean)": "clear" command may include two packets, "clear" and "clear answer _", CMS may issue a "clear" or a similar package to inform located in the same node of a DS it needs to be removed a specific file .DS send a "clear-response" or a similar package to answer: the file will eventually be removed when ready to follow the path used during the "replicate" command (available in the distribution criteria of the file. a), DS to its request «other neighboring node sends a file and its associated metadata file" clear "or equivalent commands from all the communication stations in the SCDN.

"信息(Info)":"信息"命令可以包括两个包,例如"信息" 和"信息—应答".CMS发出一个"信息"包以便向一个DS传送内容提供者元数据(与使用SCDN的内容提供者的管理有关的数据)或文件元数据.该包可以用于添加、刑除和修改特定内容提供者或文件的属性.当一个DS接收到内容提供者信息时,它修改在一个SCIWV节内的存储内容提供者元数据的表,向请求者(CMS或DS)发出"信息一应答"包,然后向它的除了请求者之外的所有邻节点发出"信息"命令.将一个包含内容提供者信息的"信息"包在整个SCDN中传播. 将一个包含文件元数据的"信息"包基于该文件的分布标准来传播。 "Information (Info)": "info" command may include two packets such as "information" and "information - Answer" .cms send a "message" to the packet metadata provider (using the SCDN transmitting content to a DS data relating to the management of the content providers) or file metadata. the packet may be used to add, modify, and other criminal attributes or content provider-specific file. when a DS receives content provider information, it modifies in a SCIWV table storage section contents provider metadata sent "a message response" packet to the requestor (CMS or the DS), and then issues "info" command to all its neighbors except the requestor. will contain a the content provider information "information" packet propagated throughout the SCDN the file metadata contains a "information" packet based on the distribution criteria of the file to spread. 当一个CMS将文件元数据的"信息"包与文件的分布标准一起发送给一个DS时,接收的DS修改其包含文件元数据的数据库,向请求者(CMS或DS)发出"信息一应答"包,然后向那些满足分布标准的邻节点(即,那些在"复制"命令期间接收到文件的分布的节点)发出"信息"包.这个过程一直持续到在满足分布标准的所有通信站中的包含文件元数据的数振库被更新. When a CMS sends the standard file metadata along with the distribution of "information" packet to a DS with a file, the received DS modifies its database containing the file metadata, sent "a response message" to the requestor (CMS or DS) package, and then to those neighbors satisfying the distribution criteria (i.e., those nodes during the "copy" command received distribution file) issued a "message" package. this process continues until at all the communication stations to meet the distribution criteria of the number of vibration library contains file metadata is updated.

"学习(learn)":"学习"命令可以由一个控制单元的学习代理发出,并且可以在将一个DS添加到SCDN并且其本地存储器需要被初始化时、或者在通信站的属性改变时、或在网络配置改变时、或在从一个故障恢复期间使用.接收"学习"命令的DS将"学习"命令传播给它的除了请求者之外的所有邻节点."学习"包携带有起始暹信站的展性.每个接收"学习"包的DS确定其通信站是否具有满足学习通信站的属性的文件,如果是,它向学习通信站中的一个DS发出"复制"来拉入相关文件. "Learning (learn)": When "learn" command may be issued by a control unit of the learning agent and may be a DS added to the SCDN and its local storage needs is initialized, or property change in the communication station, or when the network configuration changes, or during recovery from a fault. DS receiving a "learn" command to "learn" command is propagated to all its neighbors except the requestor. "learn" packet carries the initial letter Siam malleability. each receiving station "learning" DS packet communication station determines whether it has the file attributes satisfy the learning station communication, and if so, it issues "replicate" to a DS in the learning station to pull communication into documents .

"取(fetch)":"取"命令可以在以活动方式学习的同时由控制单元的学习代理来使用."取"命令可以包括两种类型的包:"取" 和"取j答".在活动学习模式中,学习代理获得要学习的媒体文件的列表、其相关内客提供者以及内容提供者的CMS的指定通信站.在此期间,这些媒体文件的文件元数据不是在本地通信站中准备好的, 因此DS不具有信息来进行搜索和下栽文件.学习代理将一个"取"包与内容的起始通信站一起发给一个本地DS. DS随后向内容提供者的CMS的指定通信站的一个DS发出一个"取一信息"包.在DS获得想要的媒体文件的文件元数据之后,它将信息存储到包含文件元数据的数据库中,并向学习代理返回"取j答".学习代理随后可以进行到发出"准备"命令来下栽媒体文件. "Take (FETCH)":. "Fetch" command may be used by the learning agent while the control unit in a learning mode active "get" command may include two types of packets: "get" and "j take the answer." in active learning mode, learning the proxy list of media files to learn, and its related within the passenger providers and content providers to specify the communication station the CMS. in the meantime, file metadata such media file is not in the local communication station in ready, DS does not have the information to make file search and tilted downward and crashed learning broker a "take" communication station starting with the content of the package sent with a local DS. DS then provide content to CMS's designation after a DS communication station issued a "take a message" package. file metadata obtain the desired DS in the media file, it will contain information stored in the database file metadata, and learn from the proxy return "take a j . "learn agent can then be sent to" prepare "command tilted downward and crashed media files.

"取—信息(fetch—info) » :"取_信息》包括两个包,"取_信息" "Take - message (fetch-info)»: "_ take information" includes two packets, "take _ Information"

和"取一信息一块".每个"取"命令在其内对一特定媒体文件和一个被许可具有该媒体文件的特定DS的标识进行编码.响应于"取"命令,一个DS向在"取"中标识的DS通信站发出"取一信息".远程DS可以以"取—信息_块"回答,"取—信息—块"包含使得本地DS将媒体、 磁道和块元数据信息保存到本地元数振存储器中所必须的信息. And "a take a message." Each "fetch" command in the identification within a particular media file and a particular DS is permitted with the encoded media file. In response to "fetch" command, a DS at the " take "DS communication station identification issued" to take a message to take "remote DS may be." - information _ block "answered," take - information - block "comprising such local DS to save the media, track, and block metadata information into information necessary for the local oscillator metadata in memory.

"停止(Stop)":"停止"命令可以包括两个包,例如"停止" 和"停止j答"."停止"命令用于关闭一个DS.当一个DS接收到一个"停止"包时,它立即以"停止—应答"进行回答,并且,根据结束需求,DS可以立即关闭或在它完成正在执行的所有任务之后关闭. "Stop (Stop)":.. "Stop" command may include two packets such as "stop" and "stop j A", "stop" command is used to close a DS when a DS receives a "stop" packet, it immediately "stop - response" to answer, and, according to the end demand, DS can turn off or shut down immediately after it has completed all the tasks being performed.

分布大有效负栽文件 Large payload file distribution plant

为了分布一个文件,内容提供者设置用于该文件的具体分布标准. 在分布菔务器(DS)将上栽的大有效负栽文件存储为块之后,内容提供者通过内容管理脹务器请求DS将文件分布到SCDN中的其他节点, 即,将内容推进到网络的边缘.分布是按照由内容提供者设置的具体分布标准进行的,可以使用前面描迷的文件分布协议(FDP).分布标准可以指定区域(例如,欧洲)、具体节点和内容提供者所期望的其他信息,以控制内容的分布.例如,分布标准可以包括在节点属性或积累属性位困中找到的信息. To distribute a file, the content provider to set a specific distribution criteria for that file. Turnip service distribution device (DS) planted on the large payload file storage plant is after block, the content provider requests the content manager service is Expansion DS will be distributed to the other nodes in the SCDN, i.e., the contents of advancing to the edge of the network. profile is in accordance with specific distribution criteria by the content provider set can describe fans front of the file distribution protocol (of FDP). distribution criteria may specify regions (e.g., Europe), specific nodes, and other information desired by the content provider, to control the distribution of content. for example, the distribution criteria may include information accumulation node attributes or attribute bits found in the trapped.

文件分布的进行如下:(l)DS通过向其邻节点发送一个通知(即, 分布请求)以宣布文件的存在和分布标准,响应于内容提供者的请求来分布一个大有效负栽文件;(2)"合格的"邻节点(即,那些满足标准的邻节点)在这个初始分布过程期间下栽文件的几个部分;(3) 通知然后从邻节点传遂给邻节点,但不传递回从其接收到分布请求的邻节点;(4)每个邻节点执行步騍2和3,直到它遣遇一个叶节点或一个"终止"节点.这样,网络中文件的分布被逐级完成. Files distributed as follows: (l) DS by sending a notice to its neighbor nodes (i.e., distribution requests) to announce the existence and the distribution criteria of the file, the content provider response to a request to distribute a large payload file plant; ( 2) "qualified" neighbors (i.e., those portions meet several criteria neighbor) during this initial distribution process tilted downward and crashed files; and (3) and then to notify neighbor nodes, but not transmitted back to the transmission from the neighbor receiving from the neighbor node distribution request; (4) each neighbor performs steps 2 and 3 Ke until it sent the case of a leaf node or a "terminating" node Thus, the distribution network file is progressively completed.

接收到一个分布请求的每一个节点将该请求传送给它的除了"请求"节点(即,它从其接收到请求的节点)之外的所有邻节点. 一个终止节点是一个该节点的属性位闺及其积累位图都与分布标准不匹配并且分布请求不能被发送给该节点的父节点的节点.对于其属性位图与该文件的内容提供者的分布标准匹配的任何节点,从具有要下栽的部分的分布路径中的最近的邻节点下栽丈件的一部分。 Each node receives a distribution request passes the request to its addition to the "requesting" node (i.e., the node from which it received the request) to all neighboring nodes other than A terminating node is an attribute, the nodes Gui and accumulation bitmaps do not match the distribution criteria and the distribution request can not be sent to a parent node of the node. for any node whose attribute matches the distribution criteria provides a bitmap of the document's content, from having to the nearest neighbors in the distribution path tilted downward and crashed portion is tilted downward and crashed portion husband member. 一旦下栽,如困7所示,DS将该文件本地存储为在不同的存储巻上展开的块,块711-713.在几个存储巻上展开该文件的过程中,输入/输出(I/O)负栽在巻之间分布,國此在内容分布和内容交付期间提高了DS的整体性能.为了本发明的目的,存储巻可以是存键设备的任何集合,例如附在服务器上的盘阵列、RAID (独立盘的冗余阵列)系统或网络附加存储器(NAS)或存储器区域网络(SAN). Once tilted downward and crashed, as shown trapped 7, the DS process is stored locally deployed in Volume different memory blocks, the block 711-713. Expand the file over several storage Volume, the input / output (I / O) Volume distribution between the negative planted, this country during content distribution and content delivery to improve the overall performance of the DS for purposes of the present invention, may be any collection of storage Volume key storage device, for example, attached to the server disk arrays, RAID (redundant array of independent disks) system or a network attached storage (NAS) or a storage area network (SAN).

困13是SCDN中的一个大有效负栽的分布的例示实施例.内容提供者使用在内容提供者的客户机系统(CPC ) 530上运行的任意内容发行和管理软件把大有效负栽文件上栽到被连接到SCDN的节点B的内容管理服务器(CMS )570中.内容提供者还把分布标准上栽到CMS 570 上.如前所迷,内容管理服务器570把所上栽的文件分成磁道文件并且为每一个磁道文件发出类似于FDP "放置"命令的命令到位于节点B中的分布服务器.在其他实施例中,CMS可以被连接到SCDN的任意节点.在节点B, DS把磁道文件分成块文件用于本地存储.文件的完全拷贝显示在用点填充的节点B上.然后,CMS向在节点B的分布服务器发出类型"分布"的FDP命令.响应于分布命令,DS向它的邻近节点A、 D和E发出一个命令来复制该内容(例如,使用FDP的"复制"命令).节点D检測复制包并判定它不应该具有该内容,因此它把复制命 Sleepy 13 is a large payload Example distribution plant illustrated embodiment SCDN the content provider using the provider of the content the client system (CPC) 530 running on the arbitrary content distribution and management software to a large payload on the plant file plummeted is connected to the node B of the SCDN content management server (CMS) 570. the content provider also plummeted distributed on the CMS 570 standard. as previously planted on the fan, the content management server 570 into the document track and issues a command similar to the file of FDP "place" command for each track file to the distribution server located in the node-B. in other embodiments, CMS may be connected to any node of the SCDN. the track files in the node B, DS into block files for local storage. complete copy of the file is displayed on the node B by filling point. then, CMS issued type "profile" of the FDP command to the distribution server at node B, respectively. in response to the distribution command, DS to its neighboring nodes a, D and E issues a command to copy the content (e.g., using the FDP "replicate" command). node D detect duplicate packets and determines that it should not have the content, it copies the command 转发到它的邻节点、节点H.节点A、 E和H检測复制包并判定它们都匹配分布标准(即,它们是"合格的"节点).当就绪时,节点A、 E和H发出命令来从SCDN中的最近的节点(例如,节点B)检索该文件的一部分.节点E和H是叶节点,西此它们不传播该复制命令.但是,节点A是带有子节点B和C的根节点.节点A不能把复制命令发回到节点B,因为它是起始节点.但是,节点A可以把复制请求发送到节点C.节点C检查分布标准并判定它是一个合格的节点,罔此它从包含所需数据的最近的节点(例如,最近的节点A、 B、 E和H)检索该文件的一部分.接着节点C把复制命令发送到节点F和G.节点F是合格的,因此它从具有该数据的最近的节点(例如, 节点B或C)检索该文件的一部分.节点G和I是不合格的,因此它们什么都不接收.节点G是终止节点,因为它的分支的积累属性不满足分布标准.这 Forwarded to its neighbor, node H. nodes A, E and H detected copy packets and determines they all match the distribution criteria (i.e., they are "qualified" nodes). When ready, nodes A, E, and H issue commands to retrieve a portion of the file from the nearest SCDN node (e.g., node B). node E and H are leaf nodes, they do not propagate the West this copy command. However, the child node with the node a is B, and C the root node. a node can not replicate command back to node B, because it is the originating node. However, node a may send a request to copy the check node C. node C distribution criteria and decides it's a qualified node indiscriminately where it contains from nearest node (e.g., nearest node a, B, E and H) required to retrieve the data portion of the file is then copied to the node C transmits the command to the nodes F and G. node F is qualified, it is therefore a part of the file retrieved from the nearest node (e.g., node B, or C) having the data node G and I are unqualified, they received nothing. node G is a terminating node because it branches the accumulation of property does not satisfy the distribution criteria. this 个初始复制过程一直持续到SCDN中的所有合格的节点至少被部分地填充.在一个或多个实施例中,大有效负栽文件的相同部分(例如,块)包含在SCDN的至少一个节点中,最好地,多个节点包含相同的部分,从而产生冗余,并且当一个或多个节点或存储巻变得不可用时防止大有效负栽文件的任何部分的丢失.例如,当一 Initial replication process continues until all the eligible SCDN nodes are at least partially filled with the embodiment, the negative part of the same large effective plant file in one or more embodiments (e.g., block) included in the at least one node of the SCDN , preferably, a plurality of nodes comprising the same portions, thereby generating redundancy, and prevent the loss of any part of the plant the large payload file when one or more nodes or storage Volume becomes unavailable. For example, when a

个存储巻(或设备)变得不可用(即,丢失)时,在那个通信站的DS 不需要采取任何专门的行动来恢复被损坏的巻的内容,罔为在那个巻中所存储的并闳此丢失的大有效负栽的郜分根据需要被自动从其它网络节点下栽以服务于用户请求.分布服务器还将一个失敗的通信站的控制信息中继到失敗的通信站的相邻通信站以防止控制命令的不适当的錄束. When storage Volume (or device) becomes unavailable (ie, lost), you do not need to take any special action in the DS communication station to restore the contents of damaged Volume, Volume indiscriminately as that stored in and Hong the missing payload large plant Gao points is automatically tilted downward and crashed requests from other network nodes as needed to serve the user. control information distribution servers also a failure of the communication station to relay a communication failure adjacent communication stations station to prevent improper recording beam control command.

在正常採作期间,分布服务器发送FDP命令,例如复制、信息、 During normal for mining, the distribution server sends FDP commands, such as copy, information,

搜索和清除命令,这些命令通过在它的控制路径中的紧邻的相邻通信 Searching and clearing adjacent communication commands that by its control path immediately

站中的其它分布服务器被转发到全部或部分两络.例如,当分布服务器接收一个FDP命令、例如复制或信息时,它基于FDP分布标准把该 Other distribution server station are forwarded to all or part of two complex. For example, when the distribution server receives a FDP commands, such as copy, or information that is based on the FDP distribution criteria to

命令发送到它的相邻DS.在当一个相邻通信站失敗时的情况中,DS Command to its neighboring DS. When a communication station in an adjacent failed case, DS

将该任务保持在它的任务队列中,并重复地重试直到该任务被成功地 The task remains in its job queue, and repeatedly retries until the job is successfully

完成.同时,DS通过把FDP命令转发到失敗的通信站的相邻DS来在 Completed. Meanwhile, by forwarding the FDP command to DS to adjacent communication station fails to DS in

失敗的通信站中临时地承担DS的角色. Failure of the communication station temporarily assume the role of the DS.

FDP使用内容提供者的分布标准来把大有效负栽文件的分布一全 FDP distribution using the content provider's standard distribution to the big payload of a whole plant file

部或部分一引导到满足内容提供者的分布标准的网络中的所有节点. 分布请求可以从树中的任意节点开始,上下周游该树,直到它到达了 A guide portion or portions of the content to meet all of the nodes of the network's distribution criteria in. Profile request can start from any node in the tree, which traveled down the tree until it reaches the

一个叶节点或到达了一个终止节点•对于具有适当属性的任何节点, 如果那些邻节点包含要被下栽的文件的部分,文件被从满足特定性能标准的最近的邻节点部分地下栽.当下栽内容时,最近的邻节点不必是虚拟树中最近的,而是在距离上最近的.这防止了自文件最初被上栽的节点的大量传输.进而,该分布的分级特性防止对初始节点(例如,节点B)周闺的网络的过度的需求.通过交付更小的块和仅仅一部分文件,该交付方法减少了两络负栽.另外,因为当分布请求到达一个"终止"节点时分布请求停止通过SCDN前进,本发明防止了不 Reached a leaf node or a terminal node • For any node with appropriate attributes, if those neighbors to be tilted downward and crashed file that contains part of the file to be tilted downward and crashed from the nearest neighbor partially meet specific performance criteria. When tilted downward and crashed when the content is not necessarily the nearest neighbor in recent virtual tree, but in the nearest. this prevents the transmission from a large number of files originally planted on the node. Furthermore, the hierarchical nature of the distribution of preventing the initial node ( For example, excess demand node B) peripheral Gui network. by delivering smaller blocks and only a part of the document, the delivery method reduces the two envelope negative plant. Further, because when the distribution request arrives at a "terminating" node distribution request stopped by SCDN proceeds, the present invention is not prevented

必要的分布请求包溢满该网络. 访河大有效负栽文件 The necessary distribution request packet overflowing the network. Visit River payload plant file

最终用户可以通过最终用户客户机系统上的一个接口、例如网络測览器来请求对大有效负栽文件(例如,电影)的访问.该请求被转发到一个将把文件提供给最终用户的合适的应用服务器(即,离最终用户更近并且具有服务于该请求的带宽的一个应用服务器),例如,用于交付大視频文件的流型服务器,或用于交付大的媒体丰富文件的 An end user via the end user interface on the client system, such as a web browser to request access to the plant a large payload file (e.g., movies). The request is forwarded to a file to the end user will be provided a suitable application server (i.e., closer to the end user and service the request having a bandwidth of an application server), e.g., for delivery of a large flow pattern video file server, or to deliver large rich media files

FTP服务器,或能把VFCS安装为它的远程文件系统以访问SCDN中的内容的任何媒体楼放器.应用服务器位于网络中,这样可以被连接到SCDN的最近的节点.SCDN节点的存储巻(即,超高速援存存储器)可以包含最终用户的所请求的文件的一些、 一点也没有、或所有的块.如果文件的類外的或者全部的内容需要在应用服务器上,SCDN 节点的VFCS代表该应用赈务器与一本地DS进行通信,以便向所有的DS的邻居发出搜索请求来定位文件的所需(未驻留)部分. FTP server, or it can VFCS mounted remote file system to access the contents of any media in SCDN floor player. The application server located in the network, which can be connected to the storage node .SCDN Volume nearest node SCDN of ( the requested file, i.e., ultra high-speed recovery storage memory) may include the end user's number, is not, or all of the blocks if the outer class files or all of the content required in the application server, v FCS representatives SCDN node the application service is in communication with the relief a local DS, in order to give the necessary (non-resident) portions of the file to locate a search request to all of the DS neighbor.

例如,假设所请求的大有效负栽文件长度为IOG字节,对应于总共20块的500M字节的存储器(即,如果每一块是500M字节).进而,假设仅仅6块这样的500M字节块本地驻留在SCDN节点中.即使仅仅所请求文件的3G字节实际上存储在SCDN节点的存储系统中, 整个文件通过VFCS "看起来"本地存在于应用服务器上.应VFCS 的请求,随着应用氣务器把该文件注入最终用户,文件的未驻留部分被从SCDN中的不同的分布服务器拉入并在本地存储.文件的各部分可能从几个分布服务器同时地检索.典型地,在SCDN上接收的数据被存储为共享的存储器(例如,本地存储巻)中的块.VFCS把存储的块文件实时地汇编并多路复合成IOG字节的文件,这样应用服务器可以使用它(例如,把文件注入最终用户). For example, assume the requested large payload file length is planted IOG bytes, corresponding to a total of 500M bytes of memory 20 (i.e., if each block is 500M bytes). Further, assume only 6 such word 500M Nodal reside locally in the SCDN node even if only 3G bytes of the requested file are actually stored in the SCDN node's storage system, the entire file by v FCS "looks" locally exist on the application server. v FCS of the request, with the application of air traffic control to the injection end user files, the file does not reside portion is pulled in and portions of the file might be retrieved from several distribution servers concurrently in the local storage of the SCDN from different distribution servers. typical , the received data is stored in the SCDN as a shared memory (e.g., Volume local storage) in the memory block .VFCS block files in real-time compilation and multiplexed into IOG byte file, so that the application server can use it (e.g., the injection end user files).

为了定位文件的未驻留部分,DS簇中的一个DS发出一个从它的相邻节点开始周游SCDN树的搜索请求.该搜索请求可以包括所请求文件的分布标准和一个使用期限计数器.例如,使用期限计数器可以规定,搜索请求只需要从请求节点开始周游SCDN的两个网络集线器. 例如,当相邻节点接收和评估该搜索请求时,节点可以使计数器减1. 当它遇到叶节点、"终止节点"或使用期限计数器为O(即,在搜索请求包括一个计数器的情况中)时搜索请求终止.在丢失的数据没有被定位并且使用期限计数器为0、即如果它被包含在搜索请求中的情况中,该技索请求继续以与初始分布过程的相反的路径来周游SCDN节点.如果文件的所请求的部分存在于一个节点中,那个节点直接回答请求DS.不具有所请求文件的任何部分的节点不回答.回答也包括发送回答及可用文件的部分的节点的性能状 To locate a file does not reside portion, a DS DS issues a cluster traveled the SCDN tree from its neighboring node search request. The search request may include the distribution criteria of the requested file and a life counter. For example, the counter may use a predetermined period, a search request from a requesting node only two traveled the SCDN network hub. For example, when the neighbor node receives and evaluates the search request, the node may decrease the counter 1. when it encounters a leaf node, "End node" or life counter is O (i.e., in the case where the search request includes a counter) is a search request terminates when no is positioned in the missing data and lifetime counter is zero, i.e., if it is included in the search request in the case, this technique cord request to proceed in the reverse path of the initial distribution process to travel the SCDN node. If the requested file is partly in a node, that node directly answer requests the DS. do not have the requested file any part of the node does not answer. answered node also includes properties like portion and transmitting the answer file available 况.当请求的DS簇从SCDN中的任jt节点接收指示它们包含部分或全部所请求的文件的回答包 Condition. When the DS cluster receives a request indicating that they contain part or all of the requested file from the SCDN nodes in the reply packet according to any jt

时,簇中的DS从那些最少拥堵的节点下栽缺少的内容并把它本地存储在分布服务器的共享的存储巻中.这样,随着应用服务器把数据提供给最终用户,分布歉务器从其它节点获得了文件的剩余部分,在应用服务器和VFCS之间的通信中没有中断. When, a cluster of those nodes from DS tilted downward and crashed the least congested lack of local content and put it in a storage Volume share distribution server. In this way, the application server with the data available to end users, distribution apology from service other nodes obtained remainder of the file, there is no interruption in communications between the application server and the VFCS.

如更早所讨论的,大有效负栽文件被拆分成各部分(例如,块文件) 并遍及SCDN分布.这样,当包含文件的各部分的节点通过搜索请求被发珑时, 一簇DS可以并行地从多个节点、尤其是从当前最少拥堵的那些节点下栽文件的部分。 E.g., large payload files plant discussed earlier split into portions (e.g., block files) and distributed throughout the SCDN. Thus, when the parts are sent Long node comprising the file search request, a cluster of DS parallel from multiple nodes, especially those parts of the file from the current node tilted downward and crashed the least congested. 基于回答包中的性能信息,起始DS簇决定从哪里下栽(即,"拉入")缺少的内容以便最小化等待时间和对其他分布服务器节点的带宽要求. Based on the performance information answer packet, it determines where to start tilted downward and crashed DS cluster (i.e., "pull") missing content so as to minimize the latency and bandwidth requirements to other distribution server nodes.

内容部分被从合适的分布服务器拉入,并由运行在一个或多个VFCS服务器上的VFCS为最终用户实时汇编.VFCS允许应用服务器查看作为一个单个的、大的虛拟文件系统存在于SCDN中的分布的存储巻. Content portion is pulled from the appropriate distribution server by running on one or more VFCS VFCS servers to end users in real-time compilation .VFCS application server is allowed to view as a single, large virtual file system is present in the SCDN Volume of distributed storage.

检索不连续的文件片段 Discontinuous segments retrieved file

从一个观点看,SCDN节点的系统存镛器中的每一个存锗的块对应于一个大有效负栽文件的一个连续片段(例如,电影的一个连续区间).例如,包含一部电影的片段,如果从第一片段到最后一个片段一个接一个地看,结果可以看到整部电影.因为相同内容部分(即, 片段)位于SCDN中的几个不同的节点上,文件的不连续的片段(例 From one point of view, the system memory is Yong SCDN node's storage germanium blocks each corresponding to a large payload file planted a consecutive segments (e.g., a continuum of film). For example, a film comprising a fragment , to see if one by one from the first segment to the last segment, the results can be seen the whole movie. since the same content portions (i.e., segments) are located on several different nodes in the SCDN, discontinuous segments file (example

如,电影的不连续的部分)可以独立地和并行地检索.这具有几个重要的壤外的作用.例如,因为DS可以从几个不同的分布服务器获得需要的内容部分,SCDN的可靠性和实用性大大地增加了.另外,最终用户可以有效地"次序颠倒地"访问大有效负栽的片段,例如,可以实现电影的快进而实际上不必下栽不被实际观看的电影的所有部分.重要地,修剪(释放由一些块使用的存储器以供其它块使用)可以基于特定的内容提供者政策在"块級别"(相对于整个"文件级別,,)上实施,例如修剪可以基于使用模式.内容的使用也可以列入块级別上. E.g., discrete portions of the film) can be retrieved independently and in parallel. This has several important effects outside the soil. For example, DS can be obtained as required from the part of the content of several different distribution servers, SCDN reliability and practicality greatly increased. in addition, end users can effectively "out of order" access to all parts of the large payload fragments planted, for example, the film can be achieved faster and thus need not actually be tilted downward and crashed not actually watch the movie importantly, pruning (released by a number of memory blocks used for other blocks used) may be provided by policy implementation in the "block level" upper (with respect to the entire "file level ,,) based on a particular content, such as trimming can based on usage patterns. the use of content can also be included in the block level.

块大小和文件分布块的大小影响内容分布和内容交付的性能.在判定块大小中考虑几 The size of the impact of content distribution and content delivery performance and file block size distribution block. Consider a few at decision block sizes

个重要的因素:1)以太网MTU (最大传输单元)大小,2)存储器的物理羊元的大小,3)传送一个块所需的时间(与网络带宽有关),以及4 )在内容交付期间响应于快进或倒带命令而跳过的最短可接受周期(这被称为最小闪烁区间). Important factors: 1) Ethernet MTU (Maximum Transmission Unit) size, 2) the size of the physical memory element sheep, 3) a time required for transmission (and the network bandwidth related) block, and 4) the content delivery during in response to fast forward or rewind command to skip the shortest acceptable period (this is called the minimum flashing interval).

在确定块大小中几个目标开始起作用. 一个目标是使MTU中的空间使用率最大化,这将使内容分布更有效.另一个目标是使分布节点的拥堵最小化.用于确定块大小的另一个重要目标是防止存储碎片, 因为碎片降低了文件系统的性能,这与实现其他目标又是一致的. Block size is determined in several goals come into play. Goal is to maximize space usage in the MTU, which would make content distribution more efficient. Another goal is to minimize congestion distribution node for determining a block size another important goal is to prevent fragmentation, since fragmentation reduces the performance of the file system, which is consistent with the other goals.

太大或太小的块大小会影响性能.例如,考虑快进命令.如果块大小太大,在下栽期间泉务器响应快进请求将是反应緩慢的,因为在服务器可以处理这样的请求前它不得不完成特定块文件的下栽.相反地,如杲块大小太小,快进到块的末端会非常快.如杲快大小在最小闪烁区间内,另一个分布服务器可能会通过检索包含目标内容的块来响应快进请求. Affect the performance of the previous block size is too large or too small. For example, consider the fast forward command. If the block size is too large, tilted downward and crashed Springs traffic during a fast forward request response reaction will be slow, since the server can handle such requests it has to complete a particular block file tilted downward and crashed. Rather, as Gao block size is too small, fast forwarding to the end of the block will be very fast. the fast Gao minimum size blinking interval, another distribution server by retrieving may comprise blocks of the target content in response to fast-forward request.

例如,基于上述标准,对于支持SCDN的物理网络所傲的假设和涉及闪烁区间的大小和最小化网络拥堵的假设, 一个实施例中的块大小例如可以是2S6K字节.对于本领域的那些普通技术人员,当一个或多个假设改变时块大小可以改变并且这个值可以是实现方式和应用专用的. For example, based on the above criteria for physical network supports SCDN proud hypothesis and assumed that the size and minimizing network congestion involves flashing interval, a block size, for example, the embodiment may be 2S6K bytes. For Those of ordinary art, when one or more of the block size may be changed to change the hypothesis and this value may be specific implementations and applications.

另外,在交付两络的存储器中的块大小不必为常数.网络中的每一个DS可以根据存储设备的具体性质和它的网络条件而具有自己的设置.每一个文件可以具有不同的块大小. Further, the block size in the memory to deliver two envelope is not necessarily constant. A DS of each network may have its own setting depending on the specific nature of the storage devices and its network condition. Each file can have different block sizes.

在初始文件分布阶段期间分布到一个合格节点的块的数量由多个因素来确定,包括从源DS(即,文件被上栽的地方)到节点的距离、 节点前端和后端带宽(当前网络条件)、预測的网络条件、节点的历史使用信息、内容提供者设定的"流行"指数、以及那个节点上对内容提供者的可用的存储器. Quantity distributed to a qualified node during the initial block of the file distribution phase is determined by a number of factors, including the source DS (i.e., where the file is planted) of the distance to the node, the node front and rear bandwidth (from current network conditions), the predicted network conditions, nodes historical usage information, content available on the "pop" index, and the node set by providers of content available memory.

可伸縮性和适应性(学习和修剪) Scalability and adaptability (learning and trim)

依据本发明的一个实施例的SCDN是高度可伸缩的.例如,当一个新的节点被加入到SCDN时,它通过采用几个不同的可适应的初始化过程中的一个来下栽它需要的初始内容.在本发明的一个实施例 SCDN according to an embodiment of the present invention is highly scalable. For example, when a new node is added to the SCDN, it is by using one of several different adaptable initialization process in that it requires an initial tilted downward and crashed to content. in one embodiment of the present invention.

中,使用了一个"自动初始化"过程.当一个节点被添加到SCDN时, 就给定了一組属性.在自动初始化过程中,节点一连接到网络,它就向它所有的邻节点发出一个FDP "学习"或类似的请求.节点将其属性在学习请求中编码.邻节点把与新节点的展性一致的内容提供给它用于下栽.然后,邻节点把新节点学习请求转发给它们的采取类似操作的所有的邻节点.这样,新节点的学习请求周游整个网络,如果网络中的所有节点具有适合于新节点的内容,则它们响应该学习请求. 这个新节点收集所有的信息,下栽需要的初始内容,现在是SCDN的一个功能单元. Using an "automatic initialization" procedure when a node is added to the SCDN, for a given a set of properties in the automatic initialization process, a node connected to the network, it sends a to all its neighbors of FDP "learn" or similar request to the node attributes coded. neighbor content consistent with the new node to malleability it tilted downward and crashed for learning request. then, the new neighbor node forwards the request to learn they take a similar operation for all neighboring nodes. Thus, the learning request new node traveled the whole network, if all nodes in the network suited to the content of the new node, which responds to the learning request. the new node collects all the information the initial content tilted downward and crashed needed now is a functional unit of the SCDN.

这个自动初始化过程的例子将在困25的情况下讨论,困25显示了添加到SCDN的新的节点J. 一旦连接到SCDN就由节点J来初始化学习请求.它向它的所有的邻节点发出包含其属性位困的学习请求. Examples of this auto-initialization process will be discussed in the case of 25 trapped, stuck 25 shows a new node is added to the SCDN SCDN J. Once connected to the node J can be initialized by the learning request it sends to all its neighboring nodes the learning request comprising attribute bits trapped.

在这个例子中,仅仅有一个直接的邻节点,节点G.当节点G接收"学习"请求时,它比较节点J的属性位困和它自己的内容的分布标准.如果它具有任何合适的供节点J下栽的内容,节点G回答节点J. 然后,节点G把学习请求转发到它的除了起始节点的节点J的所有的邻节点(即,节点I和C).请求被转发到的每一个节点同样也评估该学习请求,有条件地回答,并且根据前面所述的来传播该请求.当学习请求遇到一个叶节点或一个"终止"节点时它终止. In this example, there is only one immediate neighbor, node G. When node G receives the "learn" request, it compares node J trapped attribute bit and the distribution criteria of its own content. If it has any suitable donor content tilted downward and crashed node J, node G and J. answering node, the node G forwards the request to its addition to the study of all the neighbors (i.e., nodes I and C) node of the start node J. request is forwarded to the each node also evaluates the learn request, conditionally answer, and the request according to the propagation front. when studying the request encounters a leaf node or a "terminating" node to terminate it.

随着学习请求周游SCDN,如果所有的节点具有节点J应该具有的内容,它们响应节点J.最后,节点J收集所有的响应信息并且从这些可以最优地提供初始内容的节点下栽需要的初始内容.节点J现在可用于服务内容交付请求. With initial learning request traveled SCDN, if all nodes have node J should have contents, in response to which the node J. Finally, node J collects all the response information from these nodes and can be optimally provide desired initial content tilted downward and crashed content node J is now available for the service content delivery request.

有一些另外的情况,其中学习和适应过程可以用在本发明的其它实施例中.例如,随着大有效负栽被访问,VFCS把内容提供给应用服务器(例如,流型服务器),同时它也与分布服务器进行通信来从其他位置拉入缺少的内容部分.随着越来越多的内容部分被下栽来满足最终用户的请求,每一个内容提供者的存储空间必须被仔细地监控.基于存储器的可用性和由VFCS收集的使用信息,修剪过程可以用于移去媒体文件的特定的块.与修剪过程有关的策略应该解决:(l)什么时候修剪,(2)修剪多少,以及(3)修剪哪些块,在修剪后,服务器的存储系统可以包含整个媒体文件或由本地用户頻繁访问的文件的 There are other cases in which learning and adaptation processes may be used in other embodiments of the present invention. For example, with access to a large payload is planted, v FCS to provide content to the application server (e.g., server-flow type), and it also communicate with the distribution server from another location to pull missing content portions. as more and more content portions are tilted downward and crashed to meet the end user's request, a content provider for each storage space must be carefully monitored. memory based on the availability and use of information collected by VFCS, the pruning process can be used to remove a specific piece of media files associated with the pruning process strategy should address:. (l) when pruning, how many (2) trim, as well as ( 3) pruning which blocks, after pruning, a server's storage system may contain entire media files or frequently accessed by the local user of

不连续的片段.另外,可以向内容提供者报告:更多的存储器或更多的分布服务器、应用歉务器或VFCS服务器应该被添加到网络. . Discontinuous segments Additionally, the report may be provided to the content provider: more memory or more distribution servers, application or service apology VFCS servers should be added to the network.

具有通信站的可伸缩内容交付两络 Scalable content delivery station having two communications network

图14是可伸缩内容交付网络的另一个实施例的例示例子.SCDN 1400本质上与SCDN500(见闺5 )—样,其中由多个SCDN通倌站1410 和一个中央通信站1420替代SCDN 500的单个分布中央和单个应用服务器. FIG 14 is a SCDN500 (see Gui 5) on the exemplary embodiment of the sub-.SCDN 1400 nature scalable content delivery network to another embodiment of the - like, in which a plurality of SCDN station groom through a central communication station 1410 and 1420 replace the SCDN 500 a single distribution center and a single application server.

如在更早的实施例中,SCDN 1400的通信站被组织在一个逻辑虛拟树结构中,其中树中的每一个节点具有一組属性.这样,每一个通信站具有一个存储在节点中的属性集合,并且可以用任何方便的数据结构来表示,例如属性集合可以表示为属性位困.进一步,每个通信站(即,节点)还包含通信站的每一个子通信站的积累属性集合的表示.该表示被称为"属性的积累集合",并且任何方便的数据结构可以用于它,例如一个"积累位困",它可以被定义为来自子通信站的所有积累属性位困的"二进制OR"组合.在分布服务器簇中的分布服务器使用属性位图来分布和传送大有效负栽文件的各部分,并且它们使用合计的积累属性位困来终止不需要的消息传播.SCDN中的一个通信站被指定为"中央通信站".该中央通信站拥有一个使文本串与比特位置相匹配的属性数据库表,例如, As in the earlier embodiment, SCDN 1400 are organized in a communication station of a logical virtual tree structure in which each node in the tree has a set of properties. In this way, each communication station having a property stored in the node set, and it may be represented, for example, the attribute set can be represented as a data structure with the attribute bits trapped any convenient. further, each communication station (i.e., node) also includes information indicating attributes of each sub-accumulated communication station of communication station set this representation is referred to as "accumulated collection of attributes," and any convenient data structure can be used for it, for example, a "bit accumulation storm", which may be defined as the accumulation of all attribute bits from the sub-communication stations trapped "binary OR "combination distribution server distribution server cluster using the attribute bitmap to distribute and transfer large payload files in the plant portions, and they use the total accumulated trapped attribute bits to terminate unwanted messages in a spread .SCDN the communication station is designated as the "central communications station." the central communications station to make a text string has bit positions to match the properties of a database table, for example, 考表.中央通信站1420不必是所有内容的一个数据储存库,而是可以包含部分或全部的内容, Examination table 1420 central communication station is not necessarily a data repository for all content but may contain some or all of the content,

在本发明的一个或多个实施例中,SCDN通信站(也称为"数据中心")可以被配置为困15中所示.通信站1410包括一个分布服务器簇(DSC)1510、 一个应用服务器簇(ASC)1520、 一个控制单元(CU) 15邻、 一个共享的存储系统1S30、 一个存储开关1540和通信站内控制相关开关1515.分布厥务器簇1510使用通信链路1541和1544通过存储开关1540与存储系统1530通信.应用服务器簇1S20使用通信链路1543和1545通过存储开关1540与存储系统1530通信.控制单元1550、 分布服务器簇1510和应用服务器簇1520都通过与存储开关1540通信的通信站内控制相关开关1515通信.控制单元具有它的本地存储系统1560.在本说明书中,各种组件将在后面更详细地讨论.困16提供了困14的SCDN的一个替代示例.图16是因特网的抽 In one or more embodiments of the present invention embodiments, SCDN communication station (also referred to as "Data Center") may be configured in 15 trapped communication station 1410 includes a distribution server cluster (DSC) 1510, an application server cluster (ASC) 1520, a control unit (CU) 15 o, a shared storage system 1S30, associated control switches within a storage switch 1540 and the communication station 1515. Jue distribution service is a cluster using the communication link 1510 through 1541 and memory switch 1544 1530 1540 communicate with the storage system. 1S20 application server cluster 1543 and 1545 using the communication link via the communication memory switch 1540 and the storage system 1530. the control unit 1550, are in communication with a communication memory switch 1510 and application server cluster 1540 1520 distribution server cluster relevant switches within the control station 1515 communicates with the control unit, the various components will be discussed in more detail in its local storage system 1560. in the present specification hereinafter. trapped 16 provides an alternative exemplary SCDN trapped 14. Figure 16 is the Internet pumping

象表示.可以与困"的网络中心S05相比较的, 一般地由长途通信提 Like FIG. May be trapped "in S05 hub comparison, typically by the telecommunication extract

供者所拥有和运行的高速主干1605-1到通过同层点1610"1(和 Donor owned and operated by a high speed backbone 1605-1 to 1610 points in the same layer "1 (and

未显示的其它同层点)连接.电信访问设施1650、服务器耕作(farm) Other points in the same layer is not shown) connected to telecommunication facilities 1650 to access the server tillage (Farm)

设施1640、企业网1630和并置设施1620与一个或多个长途通信提供 Facilities 1640, 1630 and enterprise networks and Disposal Facility at 1620 with one or more long-distance communication and provide

者进行安排来获得对一个或多个主干的访问.这些主要的设施类型中的每一个一般地在各种各样的安排下向二级^^司、集团和个人提供对 Those arrangements to obtain access to one or more of the backbone of these main types of facilities generally provide each of the two ^^ Division, groups and individuals in a variety of arrangements

主干的少量的访问.(电信访问设施1650应该被解释为包括诸如提供拨号棋拟调制解调器访问的因特网服务提供者、提供电缆调制解调器访问的电缆系统头端以及提供DSL的电话公司中央局的电信变型和其他数据通信变型,) A small amount of access to the trunk. (Telecommunications access facility 1650 should be interpreted to include Internet service providers such as providing dial-up chess intends modem access, providing cable system headend cable modem access and the provision of telecommunications variants and DSL telephone company central office other data communication variants)

如困所示,袭终用户客户机系统(例如,EUC 1655和EUC 1639) 一舨地经由电信访问设施1650或经由通过企业网1630的附属部分的连接来访问或提供电子邮件、网页和其它基于因特网的资源.CPC(例如,1636) —般地通过企业网内的连接上栽内容,尽管通过电信访问设施的访问也是很普通的. As shown trapped, the passage of the end-user client systems (e.g., EUC 1655 and EUC 1639) based on access via a telecommunication facility 1650, or via connection through subsidiary portions of corporate network 1630 to access or provide email, web pages, and the other a sampan resources .CPC the Internet (for example, 1636) - a camel through the network connection within the company planted contents, although access by telecommunications access facility is also very common.

通倌站特有本地网1621可以包括一个或多个把通信站部件接口到通倌站所位于的设施内的网络的网络集线器、开关或路由器.通信站特有本地两1621的确切的配置是通信站配置的规模、对每个通信站部件预期的各种通信流量以及通信站所位于的设施的特定细节的一个函数(即,每一个特定种类的服务器的数量和存储巻的数量).虽然通信站被显示为在并罝设施1620,但这仅仅是例示性的,因为通信站可以位于较大网络中的任何位置. Groom station through a local network 1621 may include a unique hub network within one or more of the communication stations to the interface member is positioned on groom station facility, a switch or a router. The communication station-specific local exact configuration of the two communication stations 1621 sizing, the specific details of each function of a communication station the various components expected traffic and the communication station is positioned facility (i.e., the number of storage and Volume each specific type of server). Although the communication station and catching rabbits is shown in the facility 1620, this is merely exemplary, since the communication station can be located anywhere in a larger network.

分布赈务器簇 Relief Services is a cluster distribution

分布赈务器簇(鹏C)尤其提供了系统容错性和可伸縮性.困17 是分布服务器簇配置的一个示例.DSC 1510包括多个分布服务器1710>1至1710-N (合称为1710)和一个负栽均衡器1720.分布服务器1710通过开关15柳访问共享的存储系统l530中的存储巻1730-1至l"7加-V (合称为l"O)上的数据(例如,块文件).分布服务器簇中的DS (例如,DS 171CM )可以通过作为出站通信量n40的分量的请求向相邻通信站的DS发出一个请求(例如,搜索请求).类似地,相邻通信站的DS可以通过作为入站通信量1750的分量的请求向DSC 1S10中的DS发出一个请求(例如,学习请求).负栽均衡器1720把入站通信量1750分派到分布服务器1710»1至1710-N的其中之一.负栽均衡器1720可以通过与一个标准路由器的各种拓朴組合而实现为一个负栽均街路由器,并且也可以处理出站DS通信量(例如,1740). Relief distribution service is a cluster (Peng C) in particular a system fault tolerance and scalability. Trapped distribution server 17 is an example of a cluster configuration comprising a plurality of distributed servers .DSC 1510 1710>. 1 to 1710-N (collectively 1710 ) and a negative distribution plant equalizer 1720. Liu server 1710 via the switch 15 to access the shared memory of the storage system Volume of 1730-1 l530 to l "7 plus -V (collectively l" data (e.g., the O), block file). distribution server cluster DS (e.g., DS 171CM) may issue a request by the outbound traffic to the DS component n40 neighboring station a request (e.g., a search request). Similarly, the adjacent DS communication station may issue a request (e.g., to learn request) to the DSC 1S10 in the DS by requesting component 1750 as inbound traffic. negative plant equalizer 1720 inbound traffic 1750 assigned to the distribution server 1710 »1 1710-N to one of them. negative plant equalizer 1720 may be achieved by a combination of a standard router with various topology for a negative average street plant router and may also handle the DS traffic (e.g., 1740) station .

每一个DS把它的请求(搜索,学习,等等)处理为将被执行的任务.该任务在每一个对应的服务器的存储器(例如,随机存取存储器) 中的任务队列1760>1到n石0"N (合称为1760)中被存储和管理.分別地对应于任务队列1760-1到1760^N的任务队列困象177(M到1770»N Each DS to its requests (search, learn, etc.) as a processing task to be performed. This task memory corresponding to each server (e.g., random access memory) in the task queue 1760> 1 to n stone 0 "N (collectively, 1760) is stored and managed respectively correspond to the task queues 1760-1 to 1760 ^ N trapped as task queue 177 (M to 1770» N

(合称为1770)被存储在存储系统1530中的数振库中.如下面所讨论的,通信站的控制单元(CU) 1550管理存储系统1530中的这个以及多个其它数据库.注意,仅仅为了方便进行示例,包含任务队列困象1770的数据库被映射在存储巻1730上.对于那些本领域的普通技术人员来说显而易见的是,任务队列闺象数振库和其它SCDN相关的数据库可以被映射在任何可用的存储巻上. (Collectively, 1770) stored in the database is the number of transducer 1530 in the storage system. As discussed below, the control unit of the communication station (CU) 1550 to manage the storage system 1530 and a plurality of the other databases. Note that only to facilitate sample, comprising a trapped image database job queue 1770 is mapped on the memory 1730 Volume for those of ordinary skill in the art it will be apparent that as the number of vibration Gui task queue database and other SCDN-related databases may be mapped on any available storage Volume.

每一个DS在控制单元数据库中周期地登记"心跳",并且,它们周期地监祝波此的心跳.在启动(例如,加电)期间,DSC中的分布服务器投票选出一个"主服务器".例如,如果DS 1710-1缺少了h 个心跳,其中h是可配置的参数,那么假设DS1710-1不再起作用.然后,主服务器把与DS 171(M相关的任务重新分布给DSC中的有效的 Each DS in the control unit period database registered in "heartbeat", and they wish to periodically monitor the heartbeat of this wave at startup (e.g., power up), the distribution servers in a DSC vote for a "Master Server" for example, if the DS 1710-1 missing h heartbeats, where h is a configurable parameter, then it is assumed DS1710-1 no longer active. then, the master and 171 (M DS redistribution tasks related to the DSC Effective

(即,活动的)分布菔务器.如果主服务器停止工作,則所有的活动的分布服务器投票选出一个新的主服务器.新的主服务器然后重新分布前一主服务器遣留的工作,如杲有的话. (Ie, active) distribution of turnip service is. If the primary server stops working, then all of the activities of the distribution server voted for a new master. The new master server and then redistributed to stay before a primary server severance of work, such as some Gao words.

在控制单元数据库不再可访问的情况下,分布服务器正常工作,除非其工作队列被反映到每个单个分布服务器的本地存储单元.在此期间,或者DS的工作队列数据库不能被访问,或者DS崩溃.如果任一个事件发生,則服务器将最终需要被重新启动.然后将读出和重新恢复在故障之前在其本地文件系统中保存的所有工作. In the case where the control unit of the database is no longer accessible, the distribution servers function unless its work queue is reflected to each individual distribution server local storage unit. In the meantime, DS or a work queue database can not be accessed, or DS collapse. If either event occurs, the server will eventually need to be restarted. then read and saved to resume all the work before the fault in its local file system.

应用歉务器簇 Application Qianwuqicu

应用饊务器簇(ASC)在提供系统容镨和可伸缩性的同时,尤其为最终用户提供服务(例如,把一部全长度电影注入最终用户的客户机系统).困18A到18C提供了应用服务器簇的三个例示实施例.在全部的三个实施例中,ASC 1520包括负栽均衡器1820和1830、多个应用服务器1810-1到1W0-M(合称为1W0)和多个VFCS服务器1M0-1 到184CWL(合称为1840).应用服务器簇提供容错和可伸缩系统性能. 例如,如果应用服务器的其中之一失敗或如杲VFCS服务器的其中之一失敗,其它现有的应用服务器或VFCS服务器的其中之一将分別地处理这些请求.类似地,如果需要更好的系统性能,可以根据需要增加应用服务器、VFCS服务器或存储系统的存储容量. An application service San cluster (ASC) system while providing scalability and praseodymium content, in particular to provide services to end users (e.g., full-length movie to an end-user client injection system). Trapped 18A to 18C provided the application server clusters three embodiments illustrated embodiment. in all three embodiments, ASC 1520 includes an equalizer 1820 and the negative 1830 plant, a plurality of application servers 1810-1 to 1W0-M (collectively 1W0) and a plurality of VFCS server 1M0-1 to 184CWL (collectively 1840). the application server clusters provide fault-tolerant and scalable system performance. For example, if one of the one of the application servers fails or if Gao VFCS server fails, the other conventional VFCS server or one application server will handle these requests, respectively. Similarly, if the need for better system performance, the application server may be added according to need, VFCS servers or storage capacity of the storage system.

困18A是图15的应用服务器簇1520的一个例示实施例.每一个应用服务器1810通过通信站的VFCS厥务器1840的其中之一访问数据(例如,大有效负栽文件),而VFCS服务器随后通过开关1S40来访问存储系统1530中的数振.尽管负栽均衡器1820和负栽均衡器18幼在困18A中在逻辑上显示为两个独立的单元,但是它们可以是相同的物理单元.同样,应用服务器1810和VFCS服务器18糾可以包含在相同的物理单元中,从而删去负栽均衡器1S20和负栽均衡器1M0的其中之一. Trapped 18A is a diagram illustrating the application server cluster 15, an embodiment 1520 illustrated embodiment. Each application server 1810 then by VFCS Jue traffic communication station is one of access to data (e.g., large payload plant file) 1840, the VFCS Server Although negative plant equalizer 1820 and the negative immature plant equalizer 18 to display the number of vibration by 1S40 access the storage system 1530 switches 18A trapped in two logically separate units, they may be the same physical unit. Similarly, application servers 1810 and the VFCS server 18 may comprise a correction in the same physical unit, thereby deleting the negative one and negative plant equalizer 1S20 the plant equalizer 1M0.

最终用户使用在最终用户客户机上运行的标准接口(例如网络浏览器)来请求大有效负栽文件.结杲,对于该文件的服务请求被发送到ASC并到达负我均衡器1820.对于服务请求的入站数据包由负栽均衡器18加转发到应用服务器1810的其中之一,例如应用服务器1810-1 (1810-1到1810"M中的一个任意示例).应用服务器1810-1向负栽均衡器1830发出对所需数据的请求.负栽均衡器1830选择通信站的VFCS服务器1840的其中之一、例如VFCS 1840>3 ( 1840-1到184孓L 中的一个任意示例)来处理该请求并把数据包从应用服务器181(M转发到VFCS 1840-3.假设全部的所需数据存在于存储系统1530中, VFCS 1840-3通过经由开关1540访问存储系统1530中的数据来处理该请求并把数据和响应经由负栽均衡器1830发回应用服务器1S10-1. End-user using standard interfaces (eg web browser) running on the end-user client to request large payload plant file. Knot Gao, the service request for the file is sent to the ASC and to the negative I equalizer 1820 for service request inbound packet forwarding from negative 18 plant equalizer applied to one of the application server 1810, such as application servers 1810-1 (1810-1 to 1810 "M is an arbitrary example) application server 1810-1 in a negative plant equalizer 1830 issues a request for the required data. one of the negative plant equalizer 1830 v FCS to select a communication station server 1840, for example (1840-1 to an arbitrary example of 184 L relic) VFCS 1840> 3 treated the request and data packets from the application server 181 (M forwarded to VFCS 1840-3. assuming all of the required data exists in the storage system 1530, VFCS 1840-3 processes the through switch 1540 via the data access to the storage system 1530 and in response to the data request and the negative plant equalizer 1830 1S10-1 back via the application server.

当应用服务器181(M与VFCS 1840^3建立一个对话时,负栽均衡器1830连续在应用服务器181(M与VFCS 1840-3之间来回地转发数据.如果负栽均衡器1830支持先前描述的(即在背景部分中的)"粘性"特征并且该"粘性"特征被"打开",如果VFCS 1840-3保持可用(即,如果负栽均衡器1830没有把VFCS1840-3分配给另一个应用服务器),那么在当前的对话期之外,来自应用服务器181(M的数据可以继续被送往VFCS 1840-3.当VFCS 1840-3变得不可用时,负栽均衡器1830把数振包从应用服务器181(M送往另一个VFCS服务器, 例如VFCS 1840-K 1840"1到184(WVI的另一个任意示例).VFCS 1840-1 处理来自应用服务器181(M的请求并通过负栽均衡器1830向应用服务器1810^1发送响应数据包.来自应用服务器1810-1的数据包通过负栽均衡器1820被发回到客户机.就象负栽均衡器1830,负栽均衡器1820 在最终用户的客户机系统和应用服务器18 When the application server 181 (M and VFCS 1840 ^ 3 to establish a dialogue, negative planted equalizer 1830 in application server 181 (M continuous data forwarding between 1840-3 and VFCS back and forth. If the negative planted equalizer 1830 supports previously described (i.e. in the background section) "sticky" feature and the "sticky" feature is "on", another application server, if VFCS 1840-3 remains available (i.e., if the negative plant equalizer 1830 is not assigned to the VFCS1840-3 ), then the addition of the current session, from the application server 181 (M data may continue to be sent to VFCS 1840-3. when VFCS 1840-3 becomes unavailable, the negative plant equalizer 1830 from the application packet number vibrator server 181 (M v FCS sent to another server, e.g. VFCS 1840-K 1840 "1 to 184 (another arbitrary example of WVI) .VFCS 1840-1 processing request from the application server 181 (M minus the plant by the equalizer 1830 the application server sends a response packet 1810 ^ 1 data packets from application server 1810-1 via negative plant equalizer 1820 is sent back to the client. planted as negative equalizer 1830, an equalizer 1820 planted negative in the end user client systems and application servers 18 1(M之间保持一个持久的对话.负栽均衡器1820也可以提供"粘性"特征, 1 (maintaining a sustained dialogue between M. Negative plant equalizer 1820 may also provide a "sticky" feature,

当来自不同的最终用户客户机系统的新的请求到达ASC的负栽均衡器1820时,负栽均衡器1820把新的请求转发到一个可用的应用服务器,例如应用服务器1810^3(181(M到1810^M的另一个任意示例), 应用服务器1810-3处理该请求并随后通过负栽均衡器1S:30向一个通信站的VFCS服务器,例如VFCS 1840^2 (184(Kl到1840-L的另一个任意示例)进行数据请求.然后负栽均衡器1830把数据包从应用服务器1810-3转发到VFCS 1840^2, VFCS 1840-2处理来自应用服务器1810-3的请求并通过负栽均衡器1830向应用菔务器1S10O发回响应.应用服务器181(K3通过负栽均衡器1820把响应发送到新的最终用户客户机 When a new request from a different end-user client system arrives ASC negative plant equalizer 1820, a negative plant equalizer 1820 the new request to the application server available, for example, application server 1810 ^ 3 (181 (M to 1810 M ^ another arbitrary example), the application server 1810-3 processes the request and then planted by the negative equalizer 1S: 30 VFCS server to a communication station, e.g. VFCS 1840 ^ 2 (184 (Kl to 1840-L another example any) data request is then planted negative equalizer 1830 from the data packet is forwarded to the application server 1810-3 VFCS ^ 2, VFCS 1840-2 1810-3 process requests from the application server 1840 and equalized by the negative plant 1830 to apply turnip service is 1S10O response back. the application server 181 (K3 negative planted by the equalizer 1820 response to the new end-user client

system

闺18B是困15的应用服务器组1520的另一个例示实施例.困18B 中的ASC 1520的实施例包括一个或多个应用服务器181(M到1810"M (合称为1810)、具有低带宽容量的负栽均衡器1820、具有高的网络通过量的路由器182S、负栽均衡器1830(也具有低的带宽容量)、路由器1835和一个或多个VFCS服务器184(M到184(ML(合称为1840). 每一个应用服务器通过通信站的一个VFCS服务器1840访问数据(例如,大有效负栽文件),VFCS服务器1840又依次通过开关1S40访问存储系统1530中的數据.尽管它们在困18B中在逻辑上显示为两个独立的羊元,但负栽均衡器1820和负栽均衡器18:30可以是同一个物理单元,路由器1825和路由器1835可以是同一个物理单元.并且,应用服务器1810和VFCS服务器l840可以包含在同一个物埋羊元中, 从而姻去负栽均衡器1820和1830和路由器1825和1835中的一个.这样,依据本发明的一个实施例的 18B is another embodiment Gui trapped application server group 1520 15 shown embodiment. ASC Example 18B trapped in 1520 includes one or more application servers 181 (M to 1810 "M (collectively 1810), having low bandwidth plant capacity negative equalizer 1820, a high throughput of the network router 182s, negative plant equalizer 1830 (also with low bandwidth capability), router 1835 and one or more VFCS servers 184 (M to 184 (ML (co known as 1840). each application server through a v FCS server 1840 to access data (e.g., large payload files plant) communication station, v FCS server 1840 in turn access data storage system 1S40 switch 1530 by although they are trapped 18B are shown as two separate membered sheep, but the negative and the negative planted plant equalizer 1820 18:30 equalizer may be the same physical unit, the router 1825 and the router 1835 may be the same physical unit in the logic and application server l840 and VFCS server 1810 may be contained in the same sheep were buried element, whereby negative benzoin plant to the equalizer 1820 and 1830 and 1835 and a router 1825. thus, according to one embodiment of the present invention. 置删去了负载均衡器1830、路由器用服务器1810和VFCS服务器1840合并在一起, Set deleted load balancer 1830, 1810 routers with servers and server consolidation together VFCS 1840,

本实施例支持来自VFCS服务器1840和应用服务器簇的应用服务器1810的出站通信量显著地高于入站通信量的亊实.如困18B中所示,出站通信量被发送到路由器1825和珞由器1835和负栽均衡器1820 和负栽均衡器1830,而入站通信量iftJt送到负栽均衡器但不发送到路由器.通过分离入站和出站通信量,本实施例对网络性能改善作出了贡献. This embodiment supports an application server from VFCS server 1840 and application server cluster 1810 outbound traffic is significantly higher than the inbound traffic Shi solid. As shown trapped 18B, the outbound traffic is sent to Router 1825 and Luo planted by the equalizer 1835 and the negative 1820 and the negative 1830 plant equalizer, the inbound traffic to the negative iftJt plant equalizer, but not sent to the router through both inbound and outbound traffic separation, this embodiment of the network It contributed to improved performance.

一个最终用户使用在最终用户客户机上运行的标准接口(例如网络浏览器)来请求一个大有效负栽文件.结果,对于该文件的服务请求被发送到ASC并到达负栽均衡器1820.服务请求的入站数据包由负栽均衡器1820转发到一个应用服务器1810,例如应用服务器1810-1 (181(M到181(MM[的一个任意示例).应用服务器181(M向负栽均衡器1830发出对于所需数据的请求.负栽均衡器1830选择一个VFCS 服务器1840,例如VFCS 1840-1 (184(M到1840-L的一个任意示例) 来处理该请求并把数据包从应用服务器1810-1转发到VFCS 1840-1. 假设所有的所请求的数据都存在于存储系统1530,则VFCS 184(M通过经由开关1540访问存储系统1530中的数据来处理该请求并经由路由器1835把数振和一个响应发回应用服务器1810-1. End user using standard interfaces (eg web browser) running on the end-user client to request a large payload plant file. As a result, the service request for the file is sent to the ASC and to the negative planted equalizer 1820. Service Request inbound packet forwarding from negative to a plant equalizer 1820 1810 application server, such as application servers 1810-1 (181 (M to 181 (MM [an arbitrary example). the application server 181 (M minus the plant equalizer 1830 issuing a request for the desired data. negative plant equalizer 1830 v FCS to select a server 1840, for example, VFCS 1840-1 (184 (M of any sample to a 1840-L) to handle the request and the data packet from the application server 1810- 1 forwarded to VFCS 1840-1. assuming that all of the requested data exists in the storage system 1530, the VFCS 184 (M by processing the request via the router 1835 and the number of vibration data through the switch 1540 to access the storage system 1530 a response back to the application server 1810-1.

当应用服务器1810-1建立与VFCS 184(M的对话时,负栽均衡器1830继续从应用服务器181(M向VFCS 1840-1发送数据.如果负栽均衡器1830支持"粘性"特征并且该"粘性》特征被"打开",在当前对话期之外,来自应用服务器1810-1的数据可以继续被引导到VFCS 1840-1,只要VFCS 184(M保持可用(即,如果负栽均衡器1830没有把VFCS 1840-1分配給其它的应用服务器).从VFCS 184(M到应用服务器1810-1的数据流过路由器1835.路由器1835它从VFCS 1840-1接收的数据包转发到应用服务器1810-1.应用服务器1810-1通过路由器1825把数据包发送到最终用户客户机系统. When the application server 1810-1 establish VFCS 184 (M dialogue, the negative planted equalizer 1830 continues (M sends data from the application server 181 to VFCS 1840-1. If the negative planted equalizer 1830 supports "sticky" feature and the " sticky "feature is" open ", in addition to the current session of data from the application server 1810-1 may continue to be directed to VFCS 1840-1, as long as VFCS 184 (M remains available (i.e., if no negative plant equalizer 1830 the VFCS 1840-1 allocated to other application server) from VFCS 184 (M data to the application server 1810-1 flows through router 1835. router 1835 forwards it receives from VFCS 1840-1 of packet 1810-1 to the application server the application server 1810-1 sent to the end user client system 1825 via the packet router.

当来自不同的最终用户客户机的新的请求到达ASC的负栽均衡器1820时,负栽均衡器1820把新的请求转发到一个可用的应用服务器, 例如应用服务器181(^3 (1810-1到1810"M的另一个任意示例),应用服务器1810>3处理该请求并依次通过负栽均衡器1830将数据请求发给一个VFCS服务器、例如VFCS 1840-3 ( 1840-1到1840-L的另一个任意示例)向.VFCS 1840-3处理来自应用服务器1810-3的请求并通过路由器1835把数据发回到应用服务器181CK3.应用服务器1810-3通过路由器1825把响应数振发回最终用户客户机系统. When a new request from a different end-user client arrives ASC negative plant equalizer 1820, an equalizer 1820 negative plant new request to an available application server, such as application servers 181 (^ 3 (1810-1 1810 to "another arbitrary example of M), the application server 1810> 3 processes the request and passes through the negative plant equalizer 1830 v FCS data request to a server, for example, VFCS 1840-3 (1840-1 to 1840-L of another arbitrary example) to .VFCS 1840-3 1810-3 handling requests from the application server and the application server back to the router 1835 via the data 181CK3. vibrator application server 1810-3 sends back the end-user customers through a router 1825 Responses system.

困18C是困15的应用服务器簇1520的第三个例示实施例.这个与困l肪的实施例非常相似的实施例在两个重要的方面不同:(1)进和出ASC的全部与EUC有关的数振通信量都经过路由器1825,和(2 ) 应用服务器1810和VFCS服务器1840之间的全部数据通信量都经过路由器183S. 18C is trapped trapped third embodiment of the application server cluster 1520 15 This Example illustrates the embodiment of fat trapped l embodiment differs in two important embodiment is very similar to the embodiment: (1) in and out of the ASC and all EUC number of vibration have been related to the traffic router 1825, and (2) all data traffic between application servers 1810 and the VFCS server 1840 have been router 183S.

入站客户机数据包通过路由器182S流到负栽均衡器1820.然后负栽均衡器1820通过路由器1825把入站通信量调度到应用服务器.全部的出站通信量通过路由器1825流到最终用户客户机系统.到VFCS 服务器的入站通信量从路由器1835流到负栽均衡器1830,负栽均衡器1830通过路由器1835把入站通信量调度到VFCS服务器.VFCS服务器出站通信量通过珞由器1835流到相应的应用服务器.再一次,负栽均衡器1820和负栽均衡器1830可以是同一个物理单元,路由器1825 和路由器1835可以是同一个物理单元.并且,应用服务器1810和VFCS 服务器1840可以包含在同一个物理单元中,从而删去负栽均衡器1820 和1830和路由器1825和1835中的一个.这样,依据本发明的一个实施例的配置删去了负栽均衡器1830、路由器1835,并且把应用服务器1810和VFCS服务器18训合并在一起. Inbound client data packets negative plant equalizer 182S flows through the router 1820 and the negative scheduling plant equalizer 1820 through 1825 to the application server inbound traffic router all outbound traffic flows through Router 1825 end-user client system. inbound traffic flows to a negative plant v FCS server equalizer 1830 from router 1835, the equalizer 1830 planted negative VFCS .VFCS dispatched to the server station via the communication server 1835 the amount of inbound traffic from the router device Luo 1835 flows to the corresponding application server. again, the equalizer 1820 and the negative negative plant planted equalizer 1830 can be the same physical unit, the router 1825 and the router 1835 may be the same physical unit. then, the application server 1810 and VFCS server 1840 may be contained in the same physical unit, thereby deleting negative planted equalizer and routers 1835 and 1820 and one 1830 1825. thus, the equalizer omitted negative plant a 1830 depending on the configuration of the present embodiment of the invention, the router 1835 and the application server 1810 and VFCS server 18 combined training.

困18A的实施例需要最少的硬件.再者,它需要一个具有足够容量的负栽均衡器来处理入站和出站通信量簇.闺18B和困18C的实施例对于具有繁忙的出站通信量和较低的入站通信量的簇具有明显的优势.两者都需要两个小容量负栽均衡器.在图18B的实施例中,入站和出站通信量在负栽均衡器和路由器之间分开,而在困18C的实施例中,入站和出站通信量通过路由器,它把负栽均衡器用作一个资源. 对于具有繁忙的双向通信量的应用服务器簇,例如在困18B和困18C 的实施例中,要求更小容量的负栽均衡器和路由器.图1SB和闺18C 的实施例可以使用可以以透明模式运行的负栽均衡器来配置. 18A trapped embodiments require a minimum of hardware. Further, it needs to have a sufficient capacity to handle negative plant equalizer inbound and outbound traffic cluster. Gui 18B and Example 18C for trapped with heavy outbound traffic and a lower amount of clusters of inbound traffic has obvious advantages. both require two small capacity negative plant equalizer. in the embodiment of Figure 18B, the inbound and outbound traffic and negative plant equalizer separation between the router, whereas in the embodiment trapped 18C, the inbound and outbound traffic through the router, which is used as a negative equalizer a plant resource. for application server clusters with heavy two-way traffic, for example trapped 18B Example 18C trapped, the smaller capacity required negative planted Example equalizer and routers. Gui 18C and FIGS 1SB may be used to run a negative plant equalizer configured in transparent mode.

虛拟文件控制系统(VFCS)协议和VFCS服务器在前面一部分讨论了如果全部所请求的数据都存在于通信站的存储系统中时一个VFCS服务器如何处理一个应用服务器的文件请求. 现在在闺19的环境下讨论如果全部被请求的数据都不存在于通信站的存储系统时VFCS服务器做些什么, Virtual File Control System (v FCS) protocol and v FCS server front part discusses a v FCS how the server processes an application server, if all of the requested data exists in the storage system, a communication station requests for files now environment Gui 19 the VFCS server discuss what to do if all the requested data is not present in the storage system, communication stations,

假设VFCS 184(W在为应用服务器181(M处理请求.如果应用服务器1810-1需要所请求文件的凝外或全部内容,則VFCS 184©"3寻求通信站的分布服务器簇(例如,1510)中的一个分布服务器的帮助, 以检索缺少的内容.这个通信可以采用在一个包括通信站内控制相关开关1515、在负栽均衡器1720和开关1515之间的数据通信路径1901 和在VFCS服务器1840和开关1515只见的数据通信路径1910的路径上的通信站内数据通信来完成.对于本领域普通技术人员来说显而易见的是,剛刚描迷的路径仅仅是例示性的,可以使用用于完成所需通 Suppose VFCS 184 (W in the application server 181 (M processing request. If the application requires the external server 1810-1 coagulation or all of the requested file, the VFCS 184 © "3 communication station seeking distribution server cluster (e.g., 1510) help a distribution server, to retrieve the missing content. this communication may be employed includes a control related switch 1515 within the communication station, data communication path between the negative plant equalizer 1720 and the switch 1515 of 1901 and VFCS server 1840 and the communication station in the data communication path switches data communication path 1910 to 1515 saw completed to those of ordinary skill in the art it will be apparent that the path just described is merely fans illustrative embodiment, it may be used to complete the need to pass

信的许多其他装置. Letter many other devices.

VFCS 1840^3对DSC的请求被输入到负栽均衡器1720.负栽均衡器(例如,1720)然后选择一个可用的分布泉务器、例如DS 1710>2 (171(M到中的任意一个例示性的)来服务于该请求.DS 1710-2代表应用服务器181(M向其每一个相邻通信站发出一个作为出站DS通信量的分量的搜索请求,以便定位和下栽文件的所需部分.这个搜索和下栽过程已经在上面描述了. VFCS 1840 ^ 3] The DSC request is input to the negative negative planted plant equalizer 1720. Equalizer (e.g., 1720) and then select an available service profile springs, such as> (arbitrary 171 (M to 2 DS 1710 in a exemplary embodiment) to service the request on behalf of the application server .DS 1710-2 181 (M search request as a component of a traffic station DS each of its neighboring station, in order to locate the file and tilted downward and crashed to be part of. the search and tilted downward and crashed process has been described above.

当DS 1710-2从相邻通信站接收指示它们包含整个所请求文件的部分的回答包时,DSC1510中的分布歉务器1"710将从那些最不拥堵的通信站下栽缺少的内容,并经由开关1540将其本地存储在存储系统1530中.VFCS 然后通过经由开关1540访问存储系统l530中 When DS 1710-2 receives an indication from a neighboring station which comprises a reply packet portion of the entire of the requested file, DSC1510 apology distribution service 1 "710 to those from the least congested communication stations missing content tilted downward and crashed, and its local .VFCS stored in the storage system 1530 via the switch 1540 and 1540 access the storage system by the switch via l530

的数据来处理应用服务器1810-1的请求,并经由负栽均衡器18M将数据和响应发回应用服务器1810-1. A data processing request to the application server 1810-1 and 1810-1 back to the application server via the negative plant data and the equalizer in response to 18M.

SCDN通信站的内容储存库可以包括几个存储巻(参见困7).在经由VFCS 1840访问一个内容文件期间,如果VFCS检测到组成所请求内容文件的所有块文件都不能在本地得到,則它经由一个FDP准备命令通知DSC 1510以便从其他SCDN节点下栽缺少的部分.DSC 1510 中的一个DS向其邻节点中的DS发出一个FDP搜索命令,尝试从最不拥堵的通信站定位和下栽缺少的块文件.随着块文件被下栽,内容文件的元数据被更新,以便登记块文件在本地存储巻中的存在. Content repository SCDN station may include several communication storage Volume (see trapped 7). During accessing a content file via VFCS 1840, if all the blocks of the file composed v FCS detects the content file requested can not be obtained locally, it via an FDP prepare command to notify the DSC 1510 to the missing portion .DSC 1510 tilted downward and crashed with a DS in its neighbor FDP DS issues a search command from other SCDN nodes, from the communication station attempts to locate the most congested and not tilted downward and crashed the missing block file. as the block files are tilted downward and crashed, the metadata of the content file is updated to the file registration block exists in the local storage of Volume.

在本地存储巻的故障的情况下,驻留在被破坏的本地存储巻中的数据被視为好象被VFCS丢失了.因此,当一个存储巻被破坏并且请求对一个内容文件的一些块文件的访问时,VFCS将检测到该块文件是不可访问的,并经由FDP准备或类似的命令通知DSC这些块文件是缺少的.在这种情况下,DSC中的一个DS仅仅假设该块文件在本地不存在,因此它在SCDN网络中搜索缺少的块.当缺少的块被定位时,DSC 中的DS下栽缺少的块,然后相应地更新丈件元数据.随着时间的过去,在被破坏的存储巻上的块文件被自动恢复并存储在好的存储巻中. In case of failure of local storage Volume, Volume data resides in the local storage in the destruction is considered to be lost if the VFCS. Thus, when a memory is destroyed Volume content file and a request for some of the blocks of the file when accessing, v FCS detects that the block file is inaccessible, and notifies the DSC via the FDP prepare block file or a similar command is missing. in this case, a DSC of the DS is assumed that only blocks in the file local does not exist in the SCDN network it searches for the missing blocks. when the missing block is positioned, in a DSC DS tilted downward and crashed the missing blocks and then update the metadata feet member. over time, being Volume file on the memory block is automatically restored and destruction is stored in the storage well in the Volume.

为了实现在这一部分和前面的郜分中描迷的VFCS功能,VFCS可以被实现为可堆叠文件系统、使用基本的网络文件系统诸如NFS或CIFS的代理文件服务器、或存储区域网络、或直接附装的存储器,或者实现为这些方法的组合.下面描述一种这样的实现,其中,VFCS 是使用NFS协议作为基础网络文件系统协议的代理文件服务器. In order to achieve this part of the previous Gao points in the described fans v FCS function, v FCS may be implemented as a stackable file system, use the basic network file system such as NFS or CIFS proxy file server, or a storage area network, or direct attached memory means, or as a combination of these methods. One such implementation is described below, wherein, v FCS is using the NFS protocol as an underlying network file system protocol proxy file server.

VFCS的主要任务之一是在将块文件发送回一个应用服务器之前实时地重新汇编块文件,以使得应用服务器认为该文件I/O是在一单个文件上完成的.在本发明的一个实施例中,应用服务器可以使用NFS 协议来与VFCS通信.在NFS协议中,在可以远程访问文件之前,NFS 客户机需要获得一个文件系统的根的文件句柄(handle) . NFS客户机使用NFS "安装"协议来从一个远程物理文件系统获得一个"文件句柄".NFS安装协议允许服务器向一个受限制的客户机集合分发远程访问特权并执行例如允许将一个远程目录树附在一个本地文件系统上的各种搮作系统特定功能. One of the main tasks of the VFCS is to block file before sending back to a server application block file reassembled in real time, so that the application server that the file I / O is done on a single file. In one embodiment of the present invention , the application server can use the NFS protocol. in the NFS protocol, before you can remotely access files, NFS clients need to get the root file handle of a file system (handle) VFCS communicate with. NFS client uses NFS "install" agreement to get a "file handle" .NFS installation protocol from a remote server to the physical file system allows a limited set of distributed client remote access and privileges such as allowing the implementation of a remote directory tree attached to a local file system of each Li kinds of systems for specific functions.

VFCS使得应用服务器能够经由VFCS分布式虚拟文件系统查看作为一个单个的大的分布式虚拟存储系统存在于SCDN通信站内的分布 VFCS server application so as to view the distribution of a single large, distributed virtual storage system is present in the SCDN via a communication station VFCS distributed virtual file system

的存储巻,虽然VFCS实际上不向应用服务器提供对真实磁盘文件的直接(物理)访问,但每个VFCS服务器可以以一种使它所服务的应用服务器感觉到VFCS分布式虚拟文件系统是物理安装在VFCS上的方式进行一个NFS安装.为了实现这一点,在执行应用服务器的一单个文件I/O请求时,每个VFCS服务器同时用作NFS服务器和NFS客户机.当它的角色是一个NFS服务器时,每个VFCS服务器必须支持NFS安装协议,而在从通信站的存储系统设备、例如NAS设备访问文件时,它必须作为一个客户机进行一个NFS安装.因此,有在VFCS中支持的两个不同的"安装"操作来相对于应用服务器透明地实现文 Volume of storage, although VFCS not actually provide direct (physical) access to real disk files to the application server, but the server can feel each VFCS VFCS distributed virtual file system to the application server for the service it is a physical mounted on the VFCS is a NFS mount. to achieve this, when a single file I / O request to execute the application server, each VFCS server as both NFS server and NFS client when it is a role when the NFS server, each VFCS server must support the NFS mount protocol in, for example, a file access device, it must be installed from a storage system, an NFS communication station apparatus as a NAS client. Thus, there is support in the VFCS two different "installation" operation with respect to the application server transparently implemented text

件1/0. Piece 1/0.

VFCS服务器文件系统概述 VFCS server file system overview

当每个VFCS服务器在其启动过程中被初始化时, 一个VFCS安装守护程序(daemon)经由NFS安装过程来初始化一个本地例示(对每个VFCS服务器是专用的)"文件系统".文件系统不必须在物理上存在于任何附带的存储设备上.在一个例示实施例中, 一个个人文件系统存在于每个VFCS服务器的主存储器中(例如,动态随机存取存储器的一些变型).每个VFCS服务器的个人文件系统使用一个在所有VFCS服务器中都相同的"VFCS名字空间".即,每个VFCS 服务器使用一个公共分层和命名协议来参考在共享系统存储器中保存的文件.更具体地,每个VFCS服务器的个人文件系统树中的每一个文件具有一个公共的和唯一的定位符(路径名)和句柄.这个公共VFCS 名字空间允许在通信站安装的每个文件具有一个唯一的与其相联系的内容定位符(文件名).这允许同一文件名由通信站的所有VFCS服 V FCS when each server is initialized during its startup, a v FCS installation daemon (daemon) to initialize a local installation via NFS illustrated (is a dedicated server for each VFCS) "File System." File system does not have to physically present on any accompanying storage device. in one embodiment each of the embodiments VFCS server, a personal file system exists in the main memory of each VFCS server (e.g., a dynamic random access memory of some variations). embodiment shown the system uses a personal file are the same in all VFCS server "VFCS name space." That is, each VFCS server uses a common hierarchical naming protocols and reference stored in the shared memory file system. more specifically, each each VFCS server file a personal file system tree has a common and unique locator (path name) and handle. this common VFCS name space allows each communication station installed in a single file having associated therewith the content locator (filename). this allows the same file name by the communication station serving all the VFCS 器以相同的结果来使用.这样, 一个通信站内的所有VFCS服务器对安装在通信站的大有效负栽文件的整个集合具有相同的访问.只要一个应用服务器调用其VFCS客户机(即,在应用服务器上运行的VFCS访问例程)来安装所分布的虚拟VFCS,則返回通倌站的存储系统中的分布式虚拟VFCS文件系统的根文件句柄.然后使用这个根文件句柄来访问分布式虚拟VFCS中的大有效负栽文件.因为通信站的所有VFCS服务器都访问相同的VFCS名字空间,所以任何VFCS服务器都可以使用根句柄.在一个VFCS服务器崩清的情况下,(在应用服务器上运行的)VFCS客户机使用根文件句柄重试访问所请求的文件.重试请求被发送给一个不同的VFCS服务器,这对于应用服务器是透明的.文件句柄周此在所有的VFCS服务器中提供了一个持久性特征,即,它可以幸免于一个或多个单独的VFCS服务器的崩溃. Is the same result is used. Thus, an all VFCS server in a communication station large installed in a communication station payload entire set of plant file has the same access as long as an application server invokes its VFCS Client (i.e., the application running on the server VFCS access routine) to install the distributed virtual VFCS, the distributed storage system through groom station in the root file handle is returned virtual VFCS file system. this root file handle is then used to access the distributed virtual VFCS the large payload plant file. because all VFCS server communication stations have access to the same VFCS name space, so any VFCS server can use the root handle. in the case of a VFCS server collapse clear, (run on the application server ) v FCS client uses the file in the root file retry access handle the request. retry request is sent to a different v FCS server, which is transparent to the application server. filehandle circumferential this provides a durable all v FCS server characteristics, i.e., it can survive crashes of one or more individual VFCS server.

通信站控制单元和数据储存库 The communication station control unit and a data repository

图20是依据本发明的一个实施例的SCDN中的一个通信站的控制单元和数据的示例.通信站1410包括控制单元1550和数据储存库2000.在本发明的一个实施例中,控制单元1550可以是一个包括四个子系统的嵌入式系统.困20标识出一个通信站的控制单元的子系统和数据^f存库.控制羊元1550包括资源管理子系统20邻、学习子系统2060、存锗器管理子系统2070和服务管理子系统2080,合称为控制子系统.通信站内的数据储存库2000包括网络资源储存库2010、历史和统计资源储存库2030、事件资源储存库2040和内容储存库ISO. FIG 20 is an example of a control unit and data communication station SCDN in accordance with one embodiment of the present invention. The communication station 1410 includes a control unit 1550 and data repositories 2000. In one embodiment of the invention, the control unit 1550 may be an embedded system comprises four subsystems. trapped subsystem 20 identifies the data control unit and a communication station ^ f warehousing control sheep element 1550 includes o-resource management subsystem 20, subsystem 2060 learning, memory germanium is service management subsystem management subsystem 2070 and 2080, collectively known control subsystem data repository within the communication station 2000 includes a network resource repository 2010, historical and statistical resources repositories 2030, 2040 and event resource repository content storage library ISO.

控制单元的服务管理子系统2080管理通信站的DS和VFCS服务器.资源管理子系统20S0支持SCDN网络和服务配置,并且运行记录和使用数据积累成活动.学习子系统2060支持自动初始化、递增内容学习和其他用于管理和控制通信站的自适应方法.存储器管理子系统2070监視蔟存储器使用,支持内容评定和修剪,并通知CMS每个内容提供者的存储器使用.服务管理子系统2080根振阈值设置来监視和控制服务,发出SNMP (简单网络管理协议)陷阱,输出企业MIB (管理信息库),并输出历史、统计和事件数据.对于本领域普通技术人员来说显而易见的是,这里所描迷的功能是为了方便而使用的,只要产生对SCDN通信站的有效的和可靠的控制和管理,其他功能可以被添加、翔减或替代. Service management control unit manages the communication station subsystem 2080 DS and VFCS server resource management subsystem 20S0 support SCDN network and service configuration, and operating records and using data accumulated to active learning subsystem 2060 supports automatic initialization, incremental learning content and other adaptive methods for management and control of a communication station. the memory management subsystem 2070 nest memory usage monitoring, assessment and trimming support content, and notifies the CMS that each content provider's use of memory. vibrator service management subsystem 2080 the threshold is set to monitor and control services, issue SNMP (SNMP) trap, output enterprise MIB (Management Information Base), and outputs the history, statistics and event data. for those of ordinary skill in the art is obvious that here fan functions described for convenience of use, so long as a reliable and effective control and management of the SCDN station communication, other functions may be added, or alternatively Xiang Save.

通信站控制单元的控制子系统 The control station subsystem controls the communication unit

资源管理子系统2050包括一个配置服务器、 一个资源管理器和一个内容使用统计积累代理.配置服务器支持通信站的配置,即在网络资源储存库中的配置数据的初始化和维护. Resource management subsystem 2050 includes a server configuration, a resource manager and a content usage statistics accumulated proxy. Configure the server supports communication station configuration, namely initialization and maintain the configuration of the network resource repository data.

中央通信站的配置服务器允许为SCDN配置一个属性表.这个配置数据也被保存在网格资源储存库2010中.采用闺25作为示例,当一个新的通信站(例如,通信站J)被创建和配置时,通信站J的资源管理器通知其父通信站(例如,通信站G)的资源管理器和中央通信站的资源管理器来更新它们的拓朴困并积累通信站J的属性.通信站G 的资源管理器更新其部分拓朴,通知其DS该新通信站,并向其相邻通信站(例如,通信站C和I)的资源管理器发送相同的数据.在通信站J的例如两个网络集线器的邻近内的所有资源管理器更新其部分拓朴困,并通知其DS.通信站G的资源管理器然后将通信站J的属性积累到通信站G的父通信站的资源管理器.在后面的连续的积累中,通信站J的属性可以沿珞到达SCDN的根通信站. Central communication station configuration server to allow a SCDN attribute table configuration. This configuration data is also stored in the repository 2010 grid resource. Gui 25 employed as an example, when a new communication station (e.g., communication station J) is created and configuring, J communication station notifies its parent resource manager communication station (e.g., communication stations G) Explorer Explorer and central communication stations update their topology J trapped and accumulated communication station attributes. G communication station resource manager topology updating section which notifies that the new DS communication station, the communication station and its neighbor (e.g., the communication stations C and I) resource manager transmit the same data in the communication station J for example, all the resource managers in the two adjacent hub portions topology update trapped, and inform the DS. the communication station G Explorer J and the communication station attributes accumulated to the parent communication station is the communication station G resource manager in the continuous accumulation of the latter, the communication station attributes can be moved Luo J SCDN roots reach the communication station.

当中央通信站的資源管理器接收到一个新的通信站通知时,它将新通信站添加到其SCDN网络拓朴中.中央通信站保存SCDN网络拓朴的一个完全的拷贝.只有中央通信站拥有SCDN属性表:这样,其资源管理器响应关于这个表中的表目的查询. When the resource manager central communication station receives a new communication station notification, it will add the new communication station to its SCDN network topology. Central communication station to save a complete copy of the SCDN network topology. Only the central communication station has SCDN attribute table: in this way, its resource manager responds on the table purpose of this table queries.

当通信站H的属性改变时,通过一连串资源管理器,属性被沿路积累到SCDN的根通倌站.通信站H的资源管理器还根据其新属性来调用其本地学习代理以更新本地内容. When the communication station H attribute change, through a series of resource manager, the attribute is accumulated along the way to the SCDN's root pass groom station communication station H resource manager further to call in accordance with its new attributes its local Learning Agent to update the local content.

在通信站H的控制单元的IP地址、或通信站H的第一DS的IP 地址、或通信站H的DS的负栽均衡路由器的IP地址改变的情况下, 通信站H的资源管理器将通知所有相邻通信站的资源管理器来更新其网络资源储存库,这些网络资源储存库然后将这些改变通知其邻居. 改变的传播可以在两个网络集线器之后终止. A case where the IP address of the router negative equalization plant DS IP address of the control unit of the communication station H or IP address of the communication station to the first DS is H, H or a change in the communication station, the communication station the resource manager H resource manager notifies all the neighboring station updates its network resource repository, the repository of network resources and inform its neighbors of these changes. propagation changes may terminate after two hub.

一个资源管理器管理本地网络资源储存库数据,与其他资源管理器进行通信以更新数据,并将配置数据分布给本地服务器、诸如DS和VFCS泉务器. 一个内容使用统计积累代理周期性地将内容服务和内容使用积累到相关的内容管理服务器. A resource manager manages a local network resource repository data, update data and communicate with other resource managers, and distributed to the local server configuration data, such as DS and VFCS spring business is. A content usage statistics accumulated proxy periodically content services and content related to the use of accumulated content management server.

学习子系统2060提供对通信站管理和控制的自适应方法.当一个通信站的属性改变、包括一个新通信站的创建时出现学习.学习还可以用于恢复由于存储器故障而导致的内容.学习是由学习子系统中的学习代理执行的,并由资源管理器调用.在本发明的一个实施例中, 有两种不同的学习模式一主动的和被动的,内容学习包括两个阶段一发现阶段和内容修改阶段. Learning Subsystem 2060 provides when a communication station attribute change, to create a new learning occurs including communication station when the communication adaptation method of station management and control. Learning can also be used to restore the contents of the memory due to failures. Learning is performed by the learning agent in the learning subsystem, invoked by the resource manager. in one embodiment of the present invention, there are two different learning models an active and passive, content learning includes two phases finding stage and content modification phase.

在主动模式中,在发现阶段中,作为属性改变的结果, 一个学习代理在网络中的所有内容管理服务器中查询要下栽或删除的内容列表. 学习代理从内容储存库中的内容提供者数据表定位内容管理服务器. 对于一个新的通信站,学习代理从它的SCDN父节点获得内容提供者数据表,并将该表保存到新通信站的内容储存库中.在发现阶段之后的内容修改阶段中,对于要被移去的特定媒体文件,学习代理通过向一个本地DS发出移去请求来删除不需要的内容.为了添加内容,学习代理在它在发现阶段获得的内容列表中迭代,并且,对于列表中的每一个文件,它向通信站H中的DS发出一个FDP "取"或类似的请求.对于每个文件,通信站H中的一个DS向合适的内容提供者的CMS的指定通信站的DS发出一个"取一信息"或类似的包.完成这一点来检索每个文件的元数据.通信站H中的一个1 In the active mode, in the discovery phase, as a result of changes in the property, all of a learning content management server proxy in a network query to be tilted downward and crashed or delete the contents of the list. Learn proxy provider data from the content repository table location content management server for a new communication station, the learning agent obtained from its SCDN parent node data table content provider and stores the content repository table to the new communication station. after the discovery phase content submissions stage, for specific media files to be removed, learn to delete unwanted proxy by issuing a request to remove a local DS content. to add content, learning agents found it in the content list obtained in stage iteration, and , for each file in the list, it issues a of FDP "fetch" or similar request for each file, a DS communication station H is provided by the CMS suitable content of H in the communication station designated DS DS communication station issues a "take a message" or a similar package. this is done to retrieve each file's metadata. a communication station 1 in H >S然后向它的每一个相邻DS发出一个FDP搜索或类似请求以定位该文件.只要网络中的一个DS、例如通信站D中的一个DS接收到对一个文件的搜索请求,该DS 就将该请求传送到它的所有合格的邻居(例如,通信站B).如果所请求的文件存在于通信站D,則通信站D中的DS向通信站H中的DS 发出一个FDP类型的"搜索—回答".通信站H的一个DS然后发出FDP "获取"来下栽该内容. 一个搜索请求可以产生被发送到学习通信站的多个"搜索—回答"命令.每个"搜索—回答"包括所搜索通信站D的莱务和网络负栽.在通信站H的DS从最不"忙"的通信站下栽每个文件. 一个文件可以被从多个位罝并行地下栽.FDP协议包括一个优先级标签.学习代理可以通过调整准备请求中的优先级标签来控制其学习速度.学习代理还可以控制它的取请求速率.主动学习的 > S and then sent it to a DS adjacent to each of FDP search or similar request to locate the file. As long as a DS in the network, for example a DS communication station D receives a search request for a file, the DS will the request is transmitted to all its qualified neighbors (e.g., the communication station B). If the requested file exists in the communication station D, D is the communication station issues an FDP type DS to the communication station H in the DS " Search - answer. "H is a DS communication station then issues of FDP" access "to a tilted downward and crashed to the content search request may be transmitted to generate a plurality of communication station study." Search - answered. "each command" Search - answer "communication station D includes the searched service and network Levin negative plant from at least the communication station H, DS" tilted downward and crashed each file busy "communication station. a file can be tilted downward and crashed from a plurality of parallel and the location .FDP protocol includes a priority tag. learning agent prepared may be controlled by adjusting the priority tag in the request its learning speed. learning agent may also control its fetch request rate. active learning

一个优点是,内容学习表可以手动创建,以引导一个学习代理刪除或下栽内容.随着DS下栽新文件,学习代理从本地DS获得历史运行记录,并确定何时完成了学习. One advantage is that the content of the learning table can be created manually, to guide a learning agent or tilted downward and crashed delete content. With the new DS file tilted downward and crashed, learning agents obtain records from local history to run DS, and determines when to complete the study.

在被动学习棋式中,在发现阶段中,学习代理向在每一个相邻通信站的DS发出一个学习请求.学习请求包括学习通信站的旧的和新的属性.每一个通信站将学习请求转发到它自己的邻通信站,使得学习请求将在整个网络中传播.在接收到一个学习请求之后,在一给定通信站(例如,B)的一个DS检查学习通信站(例如,H)的旧的和新的属性,并从通信站B的本地存储器获得通信站H的所必须的媒体文件的列表.对于列表中的每个媒体文件,在通信站B的一个DS向在通信站H的DS发出一个FDP "复制"或类似请求.通信站H集中所必须的内容的列表,推断出哪些是它需要删除的,哪些是它需要下栽的. 在被动模式中的内容修改阶段中,基于它在发现阶段中获得的内容的列表,通信站H中的DS簇向其自身发出"移去"命令来移去不必要的内容,并向发出"复制"请求的通信站(例如B)中 In passive learning chess where, in the discovery phase, learning agency to issue a request for the DS to learn each neighboring station. The old and new learning request attributes include learning communication stations. Each station will learn communication request forwarded to its own neighbor communication station, such that the learn request would propagate throughout the network. Upon receiving a learn request, a given communication stations (e.g., B) of a DS inspection learning communication station (e.g., H) a list of necessary media files of the old and new attributes, and H is obtained from the communication station the communication station B the local memory for each media file in the list, a DS at the communication station the communication station B to H the DS issues a of FDP "replicate" or similar request communication station H centralized list of necessary content, it needs to deduce which is removed, what it needs is tilted downward and crashed to content changes in the passive mode phase, it is obtained based on the list of contents in the discovery phase, the DS cluster in the communication station H to itself issue a "remove" command to remove unnecessary content, and send a "copy" of the communication station a request (e.g., B) in DS发出"获取"命令或类似请求以下栽文件的部分. 一个DS可以使学习请求的处理让步于其他更重要的请求.学习请求可能导致从一个或多个通信站对同一媒体文件的"复制》通知的发出,通过从多个位置下栽一个文件,DS防止网络中的任何特定节点的拥堵. DS issued a "get" command or a similar request to partially tilted downward and crashed file. DS can make the process of learning a request concessions on other more important requests. Learn request could lead to "copy" of the same media files from one or more communication stations notice, tilted downward and crashed through a file from multiple locations, the DS is preventing congestion of any particular node in the network.

存储器管理子系统2070包括一个作为其主要部件的存储器管理代理.存锗器管理代理l)监祝全体本地存储器的可用性和一个内容提供者的存镛器可用性;2)调整一个文件的"流性"指数;3)为每个内容提供者确定一个存储器安全阈值;4)摔定内容修剪;5)计算修剪量;6)移去一个文件的最不可能使用的块;以及7)向内容管理服务器拫告存储器使用信息和不足警告. The memory management subsystem 2070 includes a memory management agent as its main component is germanium memory management agent l) monitoring the availability of all wish a local memory and an availability of the content provider's storage Yong;. 2) to adjust a file "fluidic "index; 3) for each content provider determines a storage safety threshold; 4) a given content fell trim; 5) is calculated trim amount; 6) removing a file block is not possible to use the most; and 7) to the content management server jobs taken report memory usage information and lack of warning.

内^H供者通过一个内容管理应用程序来设置一个文件的初始"流行"指数.这个初始值用作对要在不久的将来访问的文件的似然性的预測.DS依賴于"流行"指数以及存储器可用性、通信站位置和网络环境来对要由分块过程930 (困9)榆出到本地存储器的一个文件的哪些部分、初始块大小和块数目作出决定. Within ^ H donors to set the initial file of a "popular" index through a content management application. This initial value is used as the file to be accessed in the near future to predict the likelihood of .DS depends on the "pop" and index memory availability, and location of the communication stations to the network environment, the initial block size and number of blocks for a decision on the process of block 930 (trapped 9) Yu out which parts of a file of the local storage.

存储器管理代理使用VFCS服务器历史运行记录数据和文件元数振數据库中的数据来确定一个合理的存储器安全阁值,调整一个文件的"流行"指数,并识别出最不可能使用的块.存储器安全网值是每个内容提供者在所有时间都必须保留的空闲存储器的最小量.根据存储器可用性和DS活动,存储器管理代理为每个内容提供者确定要修剪的数振的总量,并排定对最不可能使用的块的删除. Memory management agent uses historical data VFCS server log file data and metadata vibration database to determine a reasonable value memory safety Court, adjust a file's "popular" index, and identify the least likely to use a block of memory safety net value is the minimum amount each content provider at all times must be reserved free memory. the memory availability and DS activity, memory management agent provider determines to trim the number of vibration of the total amount for each content, side by side to delete the most unlikely set of blocks used.

当对一个内容提供者的内容修剪率超过一个特定界限时,存储器管理代理发出一个RPC调用(远程过程调用)来通知内容提供者所指定到的内容管理应用服务器.内容管理应用防止内容提供者在内容提供者的存储器低时上栽任何新的内容, When the content pruning rate of a content provider exceeds a certain limit, memory management agent makes an RPC call (remote procedure call) to notify the content provider designated the content management application server content management application to prevent the content provider in the content provider's low memory when planted on any new content,

在本发明的一个实施例中,服务管理子系统20柳包括下述部件: 历史运行记录处理器、统计处理器、亊件处理器、阈值监視器、陷阱代理、SNMP (简单两络管理协议)堆栈、展示代理和服务协定政策代理.历史运行记录处理器和统计处理器从本地设备和服务器收集统计和任务/交易运行记录,并把所有的运行记录和统计信息保存在历史和统计储存库.在通信站学习的同时',历史记录处理器把所有的文件下栽记录转发到学习代理来将下栽状态通知给代理.该处理器也把记录的入站和出站数据传输信息从本地DS转发到内容提供者存储器使用表.内容使用和统计数据库也由统计处理器更新. In one embodiment of the present invention, the Liu service management subsystem 20 comprises the following components: a processor history log, statistical processor, a processor member Shi, threshold monitor, the trap agent, the SNMP (Simple Network Management Protocol two ) stack, showcase agency policies and service agreements agent history log processor and processor statistics collected statistics and task / transaction log from the local device and the server, and run all the records and statistical information stored in the repository history and statistics at the same time the communication station to learn ', history tilted downward and crashed processor records all the documents to be forwarded to the study agent tilted downward and crashed status notification to the proxy. the processor is also the inbound and outbound data transfer information recorded from local DS forwarded to the content provider using the memory table. use the content and statistical databases are also updated by the statistical processor.

亊件处理器把接收到的事件保存在亊件数据储存库2040中.存储的事件可以源自任何一个DS、 VFCS服务器、控制单元或通信站中的任何其他设备.依賴于事件级別和问趙的严重性,事件处理器可以将事件报告给陷阱代理.阈值监視器不断地将各种阁值设置与由统计处理器采集的当前读数进行比较.如果达到一个阈值,則阉值监視器将警报传送给陷阱代理.陷阱代理从亊件处理器和阉值监視器接收亊件和警报.它记录亊件和警报,并将其转送给展示代理和SNMP陷阱代理,以通知操作者.展示代理将程序目标(例如,Java小程序)分布给服务管理应用程序和最终用户的浏览器.Java小程序将阁值设置、 性能数据、网络拓朴、使用运行记录和事件从数据储存库引导给网络搮作者. Shi member processor the received event is stored in the data repository Shi member 2040. The stored event can be derived from any one DS, VFCS servers, the control unit or any other device in the communication station Depending on the event level and Q. Zhao seriousness of the event handler can report events to trap agent. threshold monitor will continue to set the value of the various Court comparison with statistics collected by the processor, the current reading. if it reaches a threshold value, then castrated value monitoring will transmit an alarm to the trap agent. trap agent receives alerts from Shi Shi member and member castration values ​​and the processor monitors. it records Shi member and alarms, display and convert it to the proxy agent and SNMP traps to inform the operator the show program will target agent (for example, Java applets) distributed to the service management applications and end-user's browser .Java applet Court set the value, performance data, network topology, the use of the log and event data from the repository guide to network Li authors.

来自历史和统计储存库和事件储存库的数据可以被推向或拉自一个展示代理.服务协定政策代理从历史和统计储存库检索出数据,并将信息供给一个服务协定政策服务器,从这里可以实施商业协定和政策(例如保证的对每个顾客的服务质量). Data from history and statistics and events repository repository can be pushed or pulled from a show proxy. Policy Agent service agreements retrieve data from repository history and statistics, and information supply a service agreement policy server, from here you can implementation of commercial agreements and policies (such as guaranteed quality of service for each customer).

通信站的两络资源储存库 Network resources of the communication station two repositories

网络资源储存库2010包含网络和通倌站配置数据.网络配置数据 Network Resource Repository 2010 contains network and station configuration data through groom network configuration data

可以包括部分或整个网络的拓朴和一个完全的属性表.这个数据由资源管理子系统2050创建和更新. It may include part or all of the network topology and a complete property sheet. This data is created and updated by the 2050 resource management subsystem.

通倌站配置数振包括所有设备的IP地址、通信站属性、积累厲性、 负栽均衡配置、数据库配置、路由配置、统计數据采集间隔、心跳间隔、本地存储巻及其配置、等等. Groom pass station is configured vibrator comprising a number of IP addresses of all devices, communication station attributes, Li accumulation of negative plant balanced configuration, database configuration, routing configuration, statistic data collection interval, heartbeat interval, local storage and Volume configuration, and the like.

通倌站的历史和统计储存库 Groom station through history and statistics repository

历史和统计储存库2030存储由服务管理子系统收集的数据.这个数据储存库包括1)设备的历史运行记录,2)服务器和网络统计,3) 内容提供者使用表,以及4)内容使用和统计表.上述数据集合包含从许多来源获得的信息,这些来源包括:DS, VFCS服务器,应用脹务器,学习子系统2060,负栽均衡器,以及存储器管理子系统2070. History and Statistics data repository 2030 stores service management subsystem collected. The data repository including 1) the historical record run equipment, 2) server and network statistics, 3) the content providers using a table, and 4) the use and content . statistics said data set includes information obtained from a number of sources, including: DS, VFCS servers, application service expansion, a learning subsystem 2060, a negative plant equalizer, and a memory management subsystem 2070.

学习子系统2060使用历史信息来解释学习的状态. Learn subsystem 2060 using historical information to explain the state of learning.

服务管理子系统2080评价统计信息,并将其与一个用户定义的资源-使用阁值进行比较,以决定是经由管理控制台向操作者发送一个通知、还是根据操作者的配置来开启服务或关闭服务. Service management subsystem 2080 evaluation statistics, and the definition of a user's resource - use Pavilion value, to determine whether to send a notification to the operator via the management console, or to turn the service on or off depending on the configuration of the operator service.

存镛器管理子系统2070使用来自DS和VFCS的历史运行记录来确定何时、何种和多少内容要修剪. Yong is memory management subsystem 2070 using historical operating records from DS and VFCS to determine when, how and how much to trim.

内容管理者使用表由存储器管理子系统2070从由服务管理子系统2080收集的数据产生.由资源管理子系统20S0提供的数据积累功能将这个相同的信息的一个拷贝积累到每个内容提供者的指定内容管理服务器. The content manager table used by the memory management subsystem 2070 generated from the data collected by the service management subsystem 2080 functions provided by the data accumulated resource management subsystem 20S0 will accumulate a copy of this same information to each content provider Specifies the content management server.

内容使用和统计表数据由应用服务器使用运行记录分析程序来收集,并由菔务管理子系统2080转送到控制单元的數据库.这个相同的信息的一个拷贝被资源管理子系统2050提供的积累功能积累到每个内W供者的指定内容管理服务器. And the content usage statistics to collect data used by the application server log analysis program, by turnip service database management subsystem 2080 is transferred to the control unit. This same information is a copy of the resource management subsystem 2050 provides the accumulation function W accumulated within each donor to specify the content management server.

通信站的事件资濂储存库 Lian-owned communication station event repository

事件资源储存库2040包括支持事件(例如,通知、警告和错谈消息)、工作队列、服务器心跳和服务器登记数据的数据集.服务器登记数据包括服务器ID、服务类型等等.服务管理子系统从包括DS、 VFCS泉务器、资源管理子系统、负栽均衡器、学习子泉统和存储器管理子系统的各种各样的本地源收集通知、警告和错误消息.通信站"亊件"被组织到不同的类别中,并根据通信站配置数据产生通知并将通知转送到管理控制台.服务管理子系统20柳依靠心跳表来确定一个服务器是否在工作.DS使用服务器登记簿、心跳表和工作队列来对其服务器簇进行负载均衡,并对容错进行相互监視。 Event resource repository 2040 includes support for events (for example, notifications, warnings and error messages to talk about), work queues, data collection server and server heartbeat registration data server registration data includes server ID, service type, etc. from the service management subsystem including DS, VFCS Springs Service, a resource management subsystem, negative planted equalizer, learning a variety of local source sub-system and memory management subsystem spring collection notice, warning and error messages. communication station "Shi pieces" are organized into different categories, and generates a notification and notifies the communication station configuration data to the management console. Liu rely on the service management subsystem 20 to determine whether a heartbeat table in the server using the server work .DS register, heartbeat table, and work queue for load balancing cluster their servers, fault-tolerant and mutual monitoring. 服务管理子系统2080 使用服务器登记簿和心跳表来监視通信站中的服务,它在电源故陣时关机,在服务器故陣时重新启动服务. Service Management Subsystem 2080 uses the server registers and heartbeat table to monitor the services in the communication station, so that when power is off the array, restart services at the server so that the array.

通信站的内容锗存库 Germanium content of the communication station warehousing

内容储存库1530包括内容提供者信息、本地驻留的媒体文件的文件分布标准、文件元数据和内容块.内容提供者信息包括内容提供者的帐户信息、分配的内容管理服务器、保留的存储器、媒体文件的数目等等.文件元数振包括媒体文件的属性、关于如何将媒体文件划分成块文件的信息以及本地存储巻中对其块文件的索引.内容储存库1530可以橫跨多个本地存储巻.例如, 一个媒体文件的内容块在多个存储巻之间分布.内容储存库也被称为共享存储系统. Content repository 1530 comprises a content provider to provide information, media files residing locally distributed file standard, the file metadata and content block content provider content information management server includes a content provider account information, the allocated memory reserved, etc. the number of media files. vibrator file metadata attributes include media file, the media file on how the file into blocks and the information in its local storage Volume index block file content repository 1530 may span multiple local Volume is stored. For example, a media file content distributed among multiple memory blocks Volume content repository is also called a shared memory system.

在共享存储系统1530中有多个存储巻(困IS).这些巻包含在困20中显示的数个重要的数据库,例如:内容提供者数据表、文件分布标准表、文件元数据数据库和内容块文件.这些数据库是独立存储的, 可以任意映射到共享存储系统1530中的任何地方.内容提供者数据表和文件分布标准表也存在于每个内容管理服务器系统中.内容管理服务器中的表只包括被分配给该CMS的内容提供者的数据.在每个通信站的内容提供者数据表包括SCDN的所有内容提供者的信息,文件分布标准表只包括复制到通信站的媒体文件. A plurality of storage Volume (trapped IS) in the shared storage system 1530. Volume of these contain several significant difficulties database 20 shown in, for example: a content provider data table, a file distribution standard table, database and file metadata content block files. these databases are stored independently and may be mapped arbitrarily anywhere in the shared storage system 1530 content provider distribution criteria data table and the file table is also present in each content management server system. content management server table includes only the contents of which are assigned to the data provider's CMS. provider data in the content table of each communication station contains information about all the content providers SCDN, the file distribution criteria table includes only the media files copied to the communication station.

只有DS、 VFCS服务器和SMS 2070可以访问存储系统.DS可以读和写共享存储系统;然而,VFCS服务器和SMS 2070只能以只读方式访问它.学习代理在学习代理的初始化阶段中在创建初始内容提供者数据表时也对存储系统进行写搮作. Only DS, VFCS and SMS 2070 server can access the storage system .DS can read and write shared storage system; however, VFCS server and SMS 2070 can only access it in read-only learning agents learn proxy initialization phase in the initial creation. content providers also wrote Li storage system for data tables.

内容提供者数据表包括诸如内容提供者帐户信息、内容提供者分配的内容管理服务器地址、保留存储器、内容提供者的政策服务器等信息. The content provider data table includes a content provider such as account information and content provided by the content management server address assigned to retain memory content provider information policy servers.

文件分布标准数振库包括对每个内容丈件的分布标准. Document distribution standard library includes several oscillator distribution criteria for each piece of content husband.

块文件包含每个内容文件的实际内容.使用一个作为内容提供者ID、媒体文件ID、磁道ID和块号的组合的散列密钥来定义块文件的路径名,以加速数据访问. Block files contain the actual content for each content file. Use as a content provider ID, media file ID, track ID hash key combinations to define the block number and the path name of the block files for accelerated data access.

文件元数据数振库 Number of vibration library file metadata

文件元数椐数据库保存与块文件有关的文件元数据,包括内容提供者ID、初始流行指数、块大小、实际使用率、媒体大小、属性、最小保留大小、访问时间印记、磁道文件索引、块文件索引、存储巻索引等等. As noted in the metadata file metadata database file retains the file-related piece, including content provider ID, initial popularity index, block size, actual usage, media size, attributes, minimum retained size, access time stamps, track file indices, block document indexing, storage Volume index and so on.

就象所表明的,文件元数据指定用于其相关联文件的块大小.文件中的所有块都是相同的大小,除了表后一块,除非文件(如果文件是一个线性文件)或最后一个磁道(如果文件是一个非线性文件)的大小恰巧是块大小的精确倍数,否則最后一块会较小。 As indicated, the file metadata specifies the block size used for its associated file. All blocks are the same size of the file, in addition to a rear table, unless the file (if the file is a linear file) or the last track (If the file is a non-linear file) size happens to be an exact multiple of the block size, otherwise the last piece will be smaller.

文件元数据还包括用于构造一个块的实际位置的信息.由于各块可以在存储系统内的多个存储设备之间分布,因此可以有多个块存储器路径根目录.在每个存储器路径根目录内, 一块的位置和文件名是确定的.创建一系列子目录来确保有限数目的文件和子目录被包含在一给定目录中. Metadata file further includes information for configuration of the actual position of a block. Since the blocks can be distributed across multiple storage devices within the storage system, so there may be a plurality of blocks of memory path to the root. Root path in each memory in the directory, a file name and location is determined. create a series of subdirectories to ensure that a limited number of files and subdirectories contained in a given directory.

分布服务器相互进行通信以及与内容管理应用程序进行通信以便 And the distribution server communicates with the content management applications to communicate with one another

传送SCDN中的大有效负栽文件.当DS添加或移去共享存储系统中的块时,DS修改文件元数据数据库中的表目.在本质上用作一个销定管理器的序列服务器可以用于同步多个DS、 VFCS服务器、存储器管理 Transmission of large payload plant SCDN file. When adding or removing shared block storage system DS, DS modify the file metadata database entry. Serves as a manager server pin given sequence can be used in nature to synchronize multiple DS, VFCS servers, storage management

器等等对文件元数据数据库的访问,以防止可能的竟争(例如,冲突) 条件.VFCS服务器使用文件元数据数据库中的信息来将适当的块汇编和多路复合成应用服务器的文件.存储器管理子系统观察可用的共享存储器、内容提供者的保留存储器和使用运行记录.当可用的存储器减少时,它启动不太流行的内容的移去,以便为更流行的和新的内容腾出空间.它通过引导DS移去它们的一些最不可能被使用的相关联的块来完成这一点.它访问文件元数据数椐库来确定要求删除多少块以及哪些块. Etc. access to the file metadata database to prevent potential competition (e.g., conflict) condition .VFCS file metadata server using the information in the database to the appropriate block assembling and multiplexed into a file server application. the memory management subsystem observe the available shared memory, memory retention and use of operating records content providers. when reducing the available memory, it starts less popular content removed to make room for the more popular and new content space. it removed some of their blocks associated with the least likely to be used by the DS to accomplish this guide. it is noted in the number to access file metadata database to determine how many blocks, and which called for the deletion blocks.

内容储存库的特定实施取决于主机环境.例如,它可以是一个传统的RBBMS(相关数据库管理系统)或可以是目录中的平面文件的一个集合.为了性能和可靠性的原因,本发明的一个实施例使用目录中的平面文件来实施内容储存库.表被实施为一个具有固定长度记录的单一文件,或被实施为对于每个记录的单独的文件.采用单一文件在读取记录的完整列表时或在记录ID未知时搜索记录时具有性能优势.采用单独文件在经由一个已知记录ID访问记录时具有性能优势.内容提供者数据表、文件分布标准表和文件元数据数据库都存储在一个其路径对于DS、 VFCS嚴务器和SMS 2070用于访问存储在其中的文件元数据的接口是已知的目录中. Certain embodiments the content repository depends on the host environment. For example, it can be a conventional RBBMS (relational database management system) or may be a collection of flat files in the directory. For performance and reliability reasons, the present invention embodiment uses flat files in directories to implement the content repository. table is implemented as a single file with fixed-length records, or as embodiments for a complete list of separate files for each record. single file reading recorded or when performance advantages when searching records when the record ID is unknown. using a separate file in a known performance advantages when accessing a recording ID via content provider data table, file table and the distribution criteria file metadata stored in a database an interface for its path DS, VFCS strict service and SMS 2070 for accessing files stored therein are known in the metadata directory.

通信站操作和数据流 Operation and data flow communication station

在闺21A到21G的环境下给出了通倌站操作和数据流的例子.结合这些附困,应用脹务器、VFCS服务器和分布服务器之间的几个交互作用将被识別如下: 21A to 21G are given in the example of the environment through the Inner groom station operation and data flow in conjunction with these difficulties attached, the application of locking works, a number of interactions between the server and the VFCS distributed servers are identified as follows:

困21A显示了一个大有效负栽文件的请求、请求到一个选定VFCS 服务器的传递、对一个DS的准备该文件的请求、从共享存储器对该丈件的初始检索.表l显示了通信站交互作用表l:困21A中的通信站交互作用标号 交互作用描述2101A 从最终用户向负栽均衡和珞由功能1820和1825输入请求2應B 请求被传送到AS<1...M> 1810内的选定AS2101C 应用服务器请求被传送到负栽均衡和路由功能1830和18352102 负栽均衡和路由功能1830和183S选择VFCS服务器中的其中之一,例如VFCS 184CM (184(Mt到1840-L中的一个任意示例),并将请求传递给它.2103 VFCS184(M经由负栽均衡器1720请求一个DS "准备"该文件2104A VFCS 184(M开始从共享存储器检索数据,困21B显示了文件数据的初始返回、请求到一个选定的DS的转发、对缺少的部分的搜索请求的发出、网络检索到的部分的存储以及使用信息的记录.表2显示了通信站交 21A shows a trapped large payload file request plant, passing the request to a selected VFCS server, a request that a document prepared for the DS, the initial retrieved from the shared memory feet member. Table l shows a communication station interaction table l: interaction trapped communication stations 21A interaction numeral 2101A is a request from an end user is described in a negative balance and planted Luo 1820 and 1825 by the functions B input 2 to be transmitted to the requesting AS <1 ... M> 1810 AS2101C selected server application request is transmitted to the negative planted balancing and routing functionality 1830 and one negative planted 18352102 balancing and routing functionality 1830 and 183S v FCS server selection wherein, for example, VFCS 184CM (184 (Mt to the 1840-L an arbitrary example), and passes the request to it .2103 VFCS184 (M 1720 via a request negative plant equalizer DS "prepare" the file 2104A VFCS 184 (M starts to retrieve data from the shared memory, difficulties 21B shows data file the initial return, forward the request to a selected DS, issue a search for the missing part of the request, the retrieved network storage portion and a recording usage information. table 2 shows the cross-station communication 作用:表2:闺21B中的通信站交互作用<table>table see original document page 58</column></row> <table>困21C显示了使用信息的检索、内容提供者存储器保留和存储器可用性数据的检索、以及由控制单元命令的和由一个选定DS实现的相关存镛審管理活动.表3显示了通信站交互作用: Effect: Table 2: Interaction Gui communication station 21B in <table> table see original document page 58 </ column> </ row> <table> trapped 21C show the use of information retrieval, content providers and availability of memory reserved memory . retrieve data, and table 3 shows the control unit and a command related to a selected memory DS achieve Yong trial management activities communication station interactions:

表3:困21C中的通信站交互作用 Table 3: Sleepy communication station in the interaction of 21C

<table>table see original document page 59</column></row> <table> <Table> table see original document page 59 </ column> </ row> <table>

困21D显示了内容提供者信息的接收和更新.表4显示了通信站交互作用:表4:闺21D中的通信站交互作用 21D shows the trapped receiving and updating the content provider information in Table 4 shows the interaction between a communication station: Table 4: Gui 21D communication station in the interaction

<table>table see original document page 60</column></row> <table> <Table> table see original document page 60 </ column> </ row> <table>

困21E显示了在分配给一个CMS的初鼓通信站中的内容部分的接收和组织.表5显示了通信站交互作用: Table 5 shows trapped 21E shows the interaction allocated to the communication station receiving and organizing content portion at the beginning a communication station drum of the CMS:

表5:闺21E中的通信站交互作用_ Table 5: Gui 21E communication station in the interaction _

<table>table see original document page 60</column></row> <table> <Table> table see original document page 60 </ column> </ row> <table>

困21F显示了分布请求的接收和复制请求的发出.表6显示了通信站交互作用:表6:闺21F中的通信站交互作用标号 交互作用描述2116A 连接到与一个CMS相联系的所分配通倌站的内容管理应用程序发出一个"分布"请求来根振FDP协议分布一个文件. 所分配通信站的DSC的负栽均衡器1720 (即,困21F)接收这个分布请求.2116B 所分配通倌站的DSC的负栽均衡器选挣一个DS,并将分布请求转发給该特定DS.2116C 在接收到分布请求之后,通信站的DS向其相邻通信站发出"复制"请求包,启动依据FDP协议的内容部分向SCDN 的平衡的传送.困21G显示了复制请求的接收和获取请求的发出.表7显示了通信站交互作用:表7:困21G中的通信站交互作用标号 交互作用描述21歸 负栽均衡器1720接收一个复制请求.2116E 通信站的DSC的负栽均衡器选择一个DS,并将复制请求转发给该特定DS.2116F DS发出一个"获取"请求 21F shows the trapped request is received and copy the requested distribution Table 6 shows the interaction between a communication station: Table 6: Interaction numeral interaction communication station described in the Inner 21F 2116A is connected to the assigned through a CMS associated with the groom contents management application station issues a "distribution" request to the root of a file distribution protocol vibration FDP communication station allocated negative plant equalizer DSC 1720 (i.e., trapped 21F) receives this distribution request through .2116B allocated groom DSC stations negative earn a plant selected from the DS equalizer, and forwards the request to the particular distribution DS.2116C after receiving the distribution request, the DS communication station to issue "replicate" packet requests the neighboring station, based on starting FDP protocol content portions to the balance of the transmission display trapped 21G SCDN receives a duplication request and requesting acquisition table 7 shows the interaction between a communication station: table 7: interaction numeral interaction communication station described in the trapped 21G 21 normalized negative plant equalizer 1720 receives a copy request .2116E negative DSC communication station to select a plant equalizer DS, and forwards the request to that particular copy DS.2116F DS issues a "get" request 来从分布路径中的多个通信站检索内容.2116G 在接收到内容之后,DS在分布数振数据库和文件元数据數据库中创建相关表目.VFCS文件系统实施和初始化每个VFCS服务器維护一个实施分布式虚拟文件系统的个人文件系统.这个分布式虚拟文件系统向一个应用服务器(AS)提供了VFCS 服务器本地保存有作为一个连续的原子整体的AS所期望的内容文件的假象,而事实上该内容一般存储在一个通信站网络中的多个巻中的块中,此外,该内容的各部分可以从SCDN的其他节点传送,每个VFCS 服务器文件系统只实现一个骨架树结构,附加的文件元数据分量在根据需要和刚好及时的基础上被添加。 To .2116G After receiving the content, DS creates relevant entries .VFCS file system implementation and initialize each VFCS server maintenance and distribution of vibration database file metadata database and retrieving content from the distribution path a plurality of communication stations implementation of a distributed virtual file system of a personal file system. this system provides a distributed virtual file server VFCS local stores as a continuous illusion contents of the file aS desired atom as a whole to an application server (aS), and the fact Usually the content is stored in a plurality Volume communication stations in the network in the block, in addition, portions of the content may be transmitted to other nodes of the SCDN from each VFCS server file system implements only a skeletal tree structure and additional file metadata component being added as needed and just in time basis. 对理解本节重要的是,这些附加的分量只在需要它们的特定VFCS服务器上添加.在本节的剩下的讨论中,从每个VFCS服务器的角度描述问趙.除非特別指出,所迷行为与任何其他VFCS服务器的存在无关. This section is important to understand that these additional components only on their specific needs VFCS server to add in the rest of the discussion in this section, from the angle of each server's description VFCS asked Zhao Unless otherwise indicated, the fans behavior has nothing to do with the presence of any other VFCS server.

分布式VFCS文件系统可以以许多种方式实施.在本发明的一个实施例中,它由一个Uuix VNODE结构表示,在该结构中, 一个节点是一个目录或一个文件.每个VNODE具有一个代表SCDN中的内容提供者或内容卖主文件的唯一的标识符. 一个特定的内容提供者文件可以只采用以其内容提供者的名字编码的文件路径来检索. Distributed VFCS File System may be implemented in a number of ways. In one embodiment of the present invention, which is represented by a Uuix VNODE structure, in this configuration, a node is a directory or a file. Each VNODE has a representative of SCDN the content provider or vendor unique identifier contents of the file. a particular content provider files using a file path can only provide the name of the coder to retrieve its contents.

VFCS服务器的个人文件系统的根目录包含每个内容提供者的名字的列表和每个内容提供者的一个子目录.每个内容提供者子目录包含该提供者的文件的列表.当它的个人文件系统被初始化时,VFCS 服务器访问内容提供者名字的列表,以便初始化内容提供者目录.然而,当VFCS服务器被初始化时,内容提供者的内容文件不被初始化. 相反,随着流行的新内容到达,在一个应用服务器从DS或序列服务器请求内容文件或通知之后它们被填充.这个过程在这里被称为"逸摔性的和动态的初始化".依据本发明,为初始化选择的特定元件和它们被初始化的时间(在初始化时或在运行时)是基于系统性能考虑而选择的. A subdirectory of the root directory of the file system VFCS personal server contains each content provider's name and a list of each content provider. Each content provider subdirectory contains a list of files that provider. When it's personal when the file system is initialized, VFCS access to the content server provides a list of those names, in order to initialize the content provider directory. However, when VFCS server is initialized, the content provider's content files are not initialized. in contrast, with the new content popular Upon arrival, the contents of the request or notification from the DS or a file server in a sequence they are filled with the application server. this process is referred to herein as "dynamic initialization and fall of Yi." according to the present invention, the particular elements selected for the initialization and time they are initialized (at initialization time or at run time) are selected based on the system's performance considerations.

VFCS服务器在其主存储器中保持有几个超高速緩存,典型的例子包括: VFCS server on which is held a few main memory cache, and typical examples include:

1. 一个用于内容提供者信息的超高速緩存. 1. a cache for information content providers.

2. —个用于内容提供者文件信息的超离速援存. 2. - super for content providers file information from the speed memory aid.

3. —个包含VFCS文件系统的框架的部分的超高速緩存(即,VFCS VNODE数据,但一舨不包含整个VNODE树). 3. - v FCS frame contains a portion of the file system cache (i.e., v FCS VNODE data, but does not include a sampan entire VNODE tree).

4. 一个用于用来提高汇编和多路复合过程的性能的"预-取"块的超高速緩存. 4 for a compound to improve the performance of the process of compiling and multiplexing "pre - take" cache block.

VFCS服务器周期性地检查内容提供者数据表和文件元数据数据库,以确定这两个中的表目是否已经改变,并且,如果是,适当地更新其超离速緩存.周期性的检查也可以由一个由DS或序列服务器启动的通知机制来实現,以便在任何表目改变时、例如在一个新文件被上栽时或在一个旧文件已经被修剪以节省存锗器空间时通知所有VFCS 服务器.当一个应用服务器尝试访问一个其文件元数据信息不在超高速緩存中的文件时,VFCS服务器确认该文件不能从SCDN访问,然后返回一个错误指示"文件未找到".在本发明的一个实施例中,VFCS可以保持有如下的超高速後存:1. 内容提供者名字超高速緩存:内容提供者名字和ID.2. 文件名字超高速緩存:文件名字、ID、磁道数目、每个磁道的块数目、媒体文件的起始位置、目前驻留于共享存储系统中的块数目.3. 文件系统超高速援存 VFCS Server periodically checks the Content Provider Data Table and the File Metadata Database to determine whether both the entry has been changed and, if so, its ultra appropriately updated. Periodic inspections from the cache may be implemented by a notification mechanism or sequence initiated by DS server, so that when any entry change, such as when a new file is planted on or in an old file has been trimmed to save memory space germanium notification when all VFCS server. when an application server attempts to access a file whose file metadata information is not in the cache in time, VFCS server confirms that the file is not accessible from the SCDN, and returns an error indication, "file not found". an embodiment of the present invention embodiments, v FCS can be maintained following the ultra-high speed memory: a content provider name cache: content provider name and the file name cache ID.2: file name, ID, number of tracks, each track the number of blocks, the start position of the media file, number of blocks currently residing in the shared storage system .3. ultrafast recovery file storage system :在一个VFCSVNODE树实施例中,树中的每个节点具有一个句柄.根形成树的"第一层",其句柄是一个常数.树的"第二层"包括内容提供者节点,内容提供者ro用作句柄.树的"笫三层"包括文件节点,文件ro用作句柄.4. 预-取块超高速緩存:已经被预-取出来用在汇编和多路复合过程中的块.在具有无限存储器的理想环境中,可以通过用支持对存储在通信站的任意文件的AS请求所必须的信息来预填充VFCS服务器的个人文件系统来最优化性能.但是,在更普通的有限存储器的情景中,如前所述,进行"选择性的和动态的初始化"和"选择性的和动态的超离速援存"来实现最佳的可能的性能.VFCS服务器在它的初始化过程期间从共享的系统存储器1530读取内容提供者数据表和文件元数振数据库并把该数振的各部分超高速緩存到它的主存储器中.在服务于AS对于特定内容提供者文件的一 : In one embodiment VFCSVNODE tree embodiment, each node in the tree has a handle "first layer" of the tree root formation, its handle is a constant "second layer" of the tree includes the content provider node, a content provider. ro who handle tree as "three Zi" includes a file node, the file handle as ro .4 pre - fetch cache block: has been pre - taken out with the compilation process and the multiplexing of blocks in an ideal environment with unlimited memory, may be pre-filled VFCS server via the information-supporting aS request to any file stored in the communication station necessary personal file system to optimize performance. However, more general limited scenario memory, as described above, a "selective and dynamic initialization" and "selective and dynamic ultra-fast recovery from memory" to achieve the best possible performance in its .VFCS server initialization process during 1530 reads the content supplied from the shared memory system by file metadata tables and the transducer portions of the database and the number of vibration of its cache to main memory at the service provider to the AS for a specific content of a document 个请求时,VFCS服务器将扩充它的个人文件系统来包括特定文件的属性.VFCS服务器从在VFCS服务器的初始化期间被超髙速援存的文件元数据获得附加文件属性信息.该附加文件特定信息被保存在VFCS服务器的个人文件系统中,直到服务器重新启动或当存储器低并且该文件已经很长一段时间没有被访问时.VFCS初始化过程困22是依据本发明的一个实施例在VFCS初始化过程期闳执行的VFCS服务器的操作的一个流程困.在这个示例中,网络附加存储器(NAS)设备用于共享的存储系统.内容提供者数据表和文件元数据数振库在採作2201中被安装以提供对它们的数据的访问.对于本示例来说,假设这些文件被安装在/mp/metadata上.为了访问在存储系统中可用的整个块集合,所有的内容巻在操作2202中被安装.对于这个解释来说,安装点是/mp/Wockfiles ,例如,/mp/blockfiles/vl , /mp/Woclrfiles/v2等. When the request, VFCS server it will expand to include a personal file system attributes of a particular file server .VFCS obtain additional file attribute information from the super Gao speed v FCS server during initialization file metadata stored aid. The additional file-specific information is stored in the personal server VFCS file system until the server restarts or when memory is low and the file has not been a long time to access a .VFCS 22 is trapped initialization process in accordance with one embodiment of the present invention VFCS initialization process of a flow of operation performed VFCS server Hong trapped. in this example, network attached storage (NAS) device for storing system shared content provider data table and the number of file metadata database transducer is mounted in mining for 2201 to provide access to their data. for this example, assume these files are mounted on / mp / metadata. in order to access the entire block is available in the set of storage system, all of the contents of the Volume are mounted in operation 2202. for this explanation, the mount points are / mp / Wockfiles, e.g., / mp / blockfiles / vl, / mp / Woclrfiles / v2 like. VFCS服务器通过把内容提供者数据表加栽到超高速緩存中(例如,搮作2203)来初始化内容提供者的名字超高速援存.内容提供者名字超高速緩存包括内容提供者名字和ID.在搮作2204 中,VFCS服务器开始建立文件系统超高速緩存.在这个步緣,被输出的文件系统的结构可以被初始化.本发明的一个实施例可以使用Unix VNODE树来表示存储器中文件系统.这样,初始化包括对输出的根貝录的根VFCS VNODE的创建和对这个目录的恒定句柄的分配.对于这个解释来说,榆出的目录为/export.然后,如操作2205中所指示的, VFCS通过创建内容提供者目录来继续建立文件系统超高速緩存,即内容提供者VNODE数据和结构被初始化.例如,内容提供者目录采用/export/cp一na迈el、 /export/cp一name2等的形式.每个目录都有一个VNODE并且每个目录的句柄为提供者的ID.例如,假设工作室 VFCS name server by the content provider data table plus plummeted cache (e.g., Li as 2203) to initialize the content provider ultra-high speed memory aid content provider name cache provider includes a content name and ID. in Li as 2204, VFCS server begins to establish a file system cache. in this step edge, the structure of the file system is output may be initialized to an embodiment of the present invention may be used in Unix VNODE tree to represent the memory file system. Thus, the initialization includes the allocation of the root recording shellfish create root VFCS VNODE output and constant handle to this directory for this explanation, the elm a directory is / export. then, as indicated in operation 2205, VFCS provided by creating a content provider directory to continue to build the file system cache that content providers VNODE data and structure are initialized. for example, a content provider directory using / export / cp a na mai el, / export / cp a name2 etc. form. each directory has a VNODE for each directory and the handle of the provider's ID. for example, assume studio

(studio)x是一个内容提供者并且它的提供者ro为"123".当一个 (Studio) x is a content provider and its provider ro "123." When a

应用程序使用它的名字访问该工作室X自录时,123的句柄被返回到该应用程序.应用程序也可以使用"123"来访问工作室X目录,在这种情况下,数字句柄123被返回.在操作2206中,VFCS初始化内容提供者文件名字超髙速援存.例如,VFCS服务器基于存储器可用性从文件元数据数据库获得每一个文件提供者的文件名和ID的一个完整的或部分的列表.VFCS服务器直到这些文件被访问后才为它们创建VNODE.在採作2207中,VFCS输出它的文件系统.这是"选择性的 The application uses its name from X to access the studio while recording, the handle 123 is returned to the application. The application can also use the "123" to access the studio X directory, in this case, the handle 123 is digital returned. in operation 2206, VFCS initializes the content provider file name super Gao speed recovery storage. For example, VFCS server availability obtain a list of each file provided by the file name and ID of a full or part from a file metadata database based on memory .VFCS server until after the files are created VNODE access to them. in the mining for 2207, VFCS output its file system. this is a "selective

和动态的超高速緩存"的一个例子,即被选择用于填充并且出于性能原因确定它们何时被填充(在初始化时还是在运行时)的超高速緩存. And example of a dynamic cache, "the choice i.e. for filling and for performance reasons, to determine when they are filled (at initialization time or at run time) in the cache.

当文件被访问时在运行期间VFCS继续建立文件系统超高速緩存.当应用服务器访问一个特定文件、例如/export/cp一nameJ/file—nameX时,服务该请求的VFCS服务器在内容提供者文件一名列表中奎找文件一名,为在两个/export/cp目录中的对应文件创建VFCS一VNODE,并为该文件获得块文件的列表.同时,VFCS 服务器通过DSC负栽均衡器向DS发出一个"准备"请求包.在DS 下栽所请求的内容部分、把这些部分存储为块、更新文件元数据数据库来反映这一点之后;并且在VFCS轮询文件元数据数据库并认识到这些块存在于存储系统中之后;VFCS服务器对其进行读取、汇编并多路复合,并将内容发送给请求的应用服务器.每一个VFCS服务器执行先行读取超高速援存来提高性能. When the file is accessed VFCS continue to build the file system during operation of cache. When the application server to access a particular file, such as / export / cp a nameJ / file-nameX, to service the request VFCS server provider files a content Kui name in the list to find a file, create a VNODE VFCS corresponding files in both / export / cp directory, file and obtain a list of blocks of the file. at the same time, the negative planted VFCS server equalizer sent by DSC to DS a "prepare" request packet content portion tilted downward and crashed in the DS requested, these are stored as part of the block, file metadata update the database to reflect this in mind;. VFCS polls the file and the metadata database and recognize that these blocks exist after the storage system; v FCS server reads it, compiled and multiplexed, and transmitted to the content server requesting the application server to perform each of the look-ahead v FCS ultrafast recovery memory to improve performance.

VFCS文件访问过程 VFCS file access process

闺23显示了在运行时间期间内执行的VFCS服务器操作.当一个应用服务器请求对一个文件的访问时,在操作2301中,VFCS首先检查来看是否有为在VFCS文件系统中的文件准备的一个节点,其中VFCS泉务器提出问趙,是否为该文件存在一个VFCS节点? Gui 23 shows VFCS server operations performed during the run time. When an application requests access to a file server, the operation in 2301, VFCS first checks to see whether there is a file in the file system VFCS prepared node, which VFCS spring service is proposed to ask Zhao, whether there is a VFCS node for the file? 如果是, VFCS服务器进行到步猓2305,否則它进行到步樣2302.在步骤2302, VFCS检測它的文件名超高速緩存来判定用于该文件的一个表目是否存在于超离速緩存中(即,相关的文件元数据是否存在?).如果没有相关的文件元数据存在,VFCS服务器进行到搮作2303并读取文件元数据数据库,把该文件的相关的文件元数据超高速緩存到文件名超高速緩存中,然后进行到採作2304.但是,如果相关的文件元数据已经存在于文件名超高速援存中,VFCS服务器立即进行到操作2304, 为该文件创建VFCS文件系统节点.此后,VFCS直接进行到操作2305, If so, the server proceeds to step v FCS Guo 2305, otherwise it proceeds to step 2302. In step comp 2302, VFCS detects that the cache file name to determine the entry for a file exists in the cache transcend (ie, whether the relevant file metadata exist?). If there are no relevant file metadata exist, VFCS server to Li for 2303 and read the file metadata database, the relevant files in the metadata cache to file name cache, and then proceed to adopt for 2304. However, if the file metadata associated with the file name already exists in the ultra-high-speed memory aid, the VFCS server 2304 operating immediately to create VFCS filesystem node for the file. Thereafter, v FCS directly proceeds to operation 2305,

在操作2305中,VFCS文件服务器检查来看在预-取(瞬时)块超高速緩存中是否有所请求文件的任何块.如果没有这样的块,VFCS 歉务器进行到操作2306,在这里它从通信站的存储系统读取与所请求的文件相联系的块,把一个"准备"请求包发送到通信站的DSC来重新填充鈇少的块,执行一个先行读取,然后进行到操作2307.但是, 如果在预-取块超高速援存中对于所请求的文件有一个或多个块,VFCS 服务器直接从操作2305进行到操作2307. In operation 2305, the file server checks to see v FCS in the pre - Take (instantaneous) whether the cache block has the requested file blocks without any such block, v FCS apology service proceeds to operation 2306, where it read from the storage system, the communication station associated with the block of the requested file, to a "prepare" request to the DSC communication station to refill package Fu small blocks, performs a read ahead, and then proceeds to operation 2307 However, if the pre - ultrafast recovery block fetch stored in one or more blocks for the requested file, v FCS server 2307 proceeds to operation 2305 from operation.

在操作2307中,VFCS服务器汇编并多路复合这些块,并把此数据转发到应用服务器.VFCS可以执行先行读取超高速緩存来提高性能.进而,VFCS服务器实现包电子欺骗(packet spoofing)来把数据直接地从后端存储器运送到前端应用程序而没有进行援冲器拷贝. In operation 2307, VFCS server compiled and multiplexing these blocks, and forward this data to the application server .VFCS may perform look-ahead caching to improve performance. Further, the server implementation packages spoofing v FCS (Packet Spoofing) to the data memory is directly transported from the rear end to the front end application program is copy no overshoot recovery.

当被初始化时,VFCS文件系统在VFCS文件系统树的前两层中包含一个根节点和内容提供者目录.VFCS文件系统树的第三层包含内容提供者文件目录并且被动态地填充.如果当应用服务器请求该文件时内容提供者的文件的元数据不在超高速緩存中,VFCS服务器必须从文件元数据数振库读取文件的元数据,超高速援存该信息,然后在VFCS 文件系统的VNODE树中为该文件创建并附加一个VNODE. —旦文件被附加到VFCS文件系统,VFCS服务器向本地DS发送一个"准务" 请求包,并执行"先行读取"操作,直到它到达文件的终端. When initialized, VFCS file system contains two v FCS before the file system tree and the root node in a content provider directory .VFCS file system tree contains the content provider file directories and is filled dynamically third layer. If and when when requesting the file content provider server application metadata file is not in the cache, the server must be read from the file v FCS metadata repository metadata file number of the transducer, the ultra high-speed recovery information is stored, and the file system in the v FCS VNODE tree and attach the file to create a VNODE -. Once the file is attached to VFCS file system, VFCS server sends a "quasi-services" to local DS request packet, and perform "look-ahead" operation until it reaches the file terminal.

VFCS服务器超高速緩存的预-取块的数量依赖于VFCS服务器中的本M镛器可用性.当存储器运行慢时VFCS文件系统的笫三层可能被修剪,预-取块超高速緩存可能被刷新.基于存储器可用性,VFCS 服务器可以超高速緩存这些目录的一个完整的或部分的集合.利用有限的存储器,它仅仅为最頻繁被访问的文件超高速緩存文件元数据, 然后基于最近使用信息来定期地更新超高速緩存. VFCS server cache pre - the number of fetch blocks is dependent on the availability of the present Yong M VFCS server memory when run slower Zi three VFCS File System may be pruned, pre - fetch cache block may be refreshed the memory availability, a collection of VFCS server cache these directories a complete or partial. with limited memory, it is only the most frequently accessed files cache file metadata, based on then, based on recent usage information on a regular basis update the cache.

为了支持特定数重的内容提供者、假定XCP、特定Y个内容提供者文件、假定YF和数个并存的用户、假定ZU,为了最佳的可能的性能每一个都具有N个预-取块,VFCS服务器被配置为使用下面的方程表示的一个存储器大小: To support a specific number of content providers weight, assuming the XCP, a specific Y number of content provider files, assuming YF and several concurrent users, ZU assumed, for the best possible performance of N each having a pre - fetch block , VFCS server is configured to use a memory size represented by the following equation:

以兆字节表示的存储器大小- Memory size in megabytes of -

VFCS泉务器OS存储器需求+ VFCS服务器运行时间存储器需求+ Springs v FCS + service memory requirement of the OS server runtime memory requirements v FCS +

(每一个内容提供者名字超高速緩存所需的存储器)+ YF* (每一个文件名超高速緩存所需的存储器)+ (1+ XCP+ YF) * ( VNODE树的大小)+ZU*N* (块大小) + (1+ XCP + YF) * (size VNODE tree) (each content provider name ultra high-speed cache memory required) + YF * (required for each file name cache memory) + ZU * N * (block size)

内容公布和管理以及其它系统相关工具 Content publishing and management systems, and other related tools

内容管理应用程序是允许如下操作的一组工具:(1) SCDN的所有者(即,服务提供者)管理内容提供者帐户和(2)内容提供者管理它们在SCDN中的文件,这些工具访问和更新分布在整个SCDN的数据.内容管理应用程序使数据或请求包被传遂到合适的SCDN通信站. Content management application that allows a set of tools as follows: (1) SCDN owner (ie, service provider) to manage content provider accounts, and (2) the content providers to manage their files SCDN of these tools access and update data distributed throughout the SCDN. contents management application or a request packet so that the data is transferred to the appropriate SCDN then the communication station.

内容管理应用程序在客户机/服务器范例中执行,即,在内容提供者客户机系统上运行的客户机用于调用在(1)内容管理脹务器(CMS) 或(2) CMS和使用例如Java小程序或servkts的程序目标的内容提供者客户机系统的一个组合上;或者在其任意组合上执行的一个内容管理应用程序. Content management application in the client / server paradigm is performed, i.e., to provide the customer running on client system call for machines (1) Content Manager Service expansion device (CMS) or (2) CMS using, for example, and the content a combination of program objectives on the content of Java applet or servkts provider client system; or a content management application executing on any combination thereof.

在SCDN中可能有一个或多个CMS.除典型的资源例如处理器和存储器外,每一个CMS都具有它自己的本地存储设备,在其上存储有各种内容管理相关数据库. In SCDN may have one or more CMS. In addition to typical resources such as a processor and memory, each CMS has its own local storage devices, various content management related databases stored thereon.

每一个内容提供者使用一个特定的CMS.服务提供者把由内容提供者使用的CMS分配给一个特定的通信站.这意味着被分配的通信站处理由那个内容提供者执行内容管理应用程序而产生的CMS请求.在一个实施例中,所分配通信站的DSC中的一个DS被它的负栽均衡器逸择来处理CMS的请求.除了存储在CMS的本地存储设备上的与内容管理应用有关的数据外,其他这样的数椐被选择性地存储在通信站的存储系统中或者通信站的控制单元的本地存储设备中. Each content provider to use a specific CMS. The service provider used by the content provider CMS assigned to a particular communication station. This means that the assigned communication station processing is provided by the content provider and content management applications to perform CMS requests generated. in one embodiment, requests a DS DSC communication station is its negative plant equalizer Yi Optional processed CMS allocated in addition to the content management application is stored on a local storage device the CMS For external data, a number of other such devices noted in the control unit is stored locally in a storage system or the communication station the communication station to selectively store.

每一个内容提供者可以仅仅被分配给一个CMS:但是,多于一个内容提供者可以被分配给同一个CMS.而且, 一个CMS可以仅仅被分配给其中内容提供者可以把文件上栽到SCDN中的一个通信站.例如,当一个内容提供者被分配给一个CMS时,CMS使用FDP "信息" 包通过它的所分配的DS把内容提供者信息和CMS的IP地址传播到整个SCDN. —个内容使用统计积累客户机(相关通信站的资源管理子系统的一个部件)连接到CMS,以便为每一个内容提供者积累存储器和内客使用信息.内容管理服务器也使用FDP "信息"包来传播内容提供者数据表中的改变. Each content provider can only be assigned to a CMS:. However, more than a content provider can be assigned to one and the same CMS, a CMS can only be assigned to which content providers can put in the file plummeted SCDN . a communication station for example, when a content provider is assigned to a CMS, CMS use of FDP "info" packet by the DS of its assigned to spread the content provider information and the CMS IP address to the entire SCDN -. a content usage statistic accumulation client (a member resource management communication station subsystem) connected to the CMS, and to provide storage and accumulation within the passenger who uses the information for each content. content management server also uses of FDP "info" packet to propagate the content provider altered data in the table.

各种内容管理应用程序用于输入和管理内容提供者信息、内容的元信息和实际内容.应用程序也用于检索存储系统和内容的使用信息和性能统计. A variety of content management applications used to enter and manage content provider information, meta-information and the actual content of the application is also used to use to retrieve the contents of the storage system and the content of information and performance statistics.

内容管理应用程序工具 Content management application tool

内容管理应用程序包括位于CPC、 CMS和SCDN通信站上的工具.这些工具可以是程序目标的形式,例如,Java小程序.下面将参照困20来展示一些典型的工具. Contents management application includes a tool on the CPC, CMS and SCDN communication station. These tools may be in the form of program object, e.g., Java applets. Trapped below with reference to 20 show some typical tools.

内容提供者管理客户机是为内容提供者和搮作者提供用户界面以更新存储在CMS系统中的内容提供者表中的内容提供者信息的工具. The content provider management client is a provider of content and Li provides the user interface to update the content stored in the CMS system offers content providers the tools of the table's information.

内容提供者管理厥务器是管理内容提供者数据表并通过它的所分配通信站中的DS把信息包发送到SCDN的工具.DS在通信站的存储系统中创建它自己的内容提供者数据表.内容管理客户机是为内容提供者提供用户界面以上栽和翔除它们的内容、检查存储器使用、保留存储器、检查内容使用等的工具. Jue content provider management tool to manage traffic is a data table and the content provider via its assigned DS communication station transmits the packet to the SCDN .DS creates its own content provider in a data storage system, a communication station table content management client is a provider provides more user interface planted Xiang and their contents, to check memory usage, memory retention, check the contents of the use of such tools in addition to the content.

内容管理服务器是作为内容管理客户机工具和分布服务器之间的 Content Management Server is as between content management and distribution of client server tools

桥乘的工具,它使用FDP协议(例如,使用"放置"、"分布"和"清除"命令)与其所分配通信站中的DS通信.当内容管理客户机工具上栽一个新文件时,内容管理服务器使用FDP "放置"和"分布"包通过它的所分配的DS把该文件分布(即,插入)到SCDN中.内容管理服务器把内容提供者信息和内容分布标准存储在它的数据库中.在内容被分布到SCDN的同时,在处理分布请求中涉及的DS把与这些文件和它们的组成部分有关的信息存储在文件分布标准数据库和文件元数据数据库中.这些文件和它们的組成部分块也由在处理该请求中涉及的DS存储在SCDN中分布的内容储存库中.内容管理服务器使用FDP "清除"包来从SCDN删除一个文件.当SCDN中的DS接收到"清除"包时,它从文件分布标准数据库、文件元数振数据库删去相关信息,并从内容储存库中姻去实际的内容块. Bridge take the tool that uses FDP protocol (for example, using the "Place", "distribution" and "clear" command) to communicate with its assigned communication station in DS. When the content management client tools planted a new file, the contents of management server uses of FDP "placement" and "profile" assigned packet through its DS to the document distribution (i.e., inserted) into the SCDN the content management server to the content provider information, and content distribution criteria stored in its database at the same time the content is distributed to the SCDN, the processing distribution request DS related to these files and their part store information about the document distribution database and the file metadata database. these files and their part when the content repository block is also a DS storage involved in processing the request distribution of the SCDN in the content management server uses of FDP "clean" packet to remove a file from an SCDN. when SCDN in DS receiving the "Clear" package it distribution criteria from a file database, a file metadata database by deleting vibration-related information, and benzoin from the content repository to the actual content blocks.

内容使用统计积累客户机是在每一个通信站的控制单元中实现的工具.它把内容使用信息从历史和统计储存库转发到分配了内容提供者的CMS系统上的内容使用统计积累服务器工具.内容使用统计积累服务器工具接收使用和统计数据并把这些数据存储在使用和统计数据库中. Content usage statistics accumulated client is a tool implemented in the control unit of each communication station in which the contents using the contents of the information forwarded from history and statistics repository to the provider assigned content of CMS systems using statistical accumulation server tools. content usage statistics accumulated server receives the use of tools and statistical data and the use of these statistics and data stored in the database.

所分配通信站的控制单元的存储器管理子系统周期性地把内容提供者存镛器使用从内容数据储存库转发到内容提供者存储器工具代理.内容提供者存储器工具代理仅仅接收分配给它的内容管理服务器的内容提供者的存镣器使用,它把存储器使用数据存储在内容提供者存储器使用表中. The control unit of the communication station allocated memory management subsystem periodically the content provider uses to forward Yong memory from the content data repository to the Content Provider memory Agents. Agents memory content provider receives only contents allocated to it content provider management server using memory shackles, which use the data storage memory provided by the memory table used in the content.

收费工具和统计工具用来输出来自使用和统计数据库和使用和统计数据库的信息,这样服务和内容提供者都可以定制发票声明和与其顾客的其他通信. Fees tools and statistical tools used to output and use statistical databases and the use of information and statistics from the database, so that the service and content providers can customize invoice statements and other communications with their customers.

数据库内容的示例 Examples of database content

图24是田20的内容管理和使用数据库的内容的示例.在该示例中使用了两个CMS、 CMS 2410和CMS 2411,三个通信站、通信站2401、 通信站2柳2和通信站2403以及两个内容提供者CPC-1和CPC-2.在这个例子中,将做下迷假设: FIG 24 is an example of the contents and content management database 20 in the field. Two CMS used in this example, CMS 2410 and CMS 2411 of, three communication stations, communication station 2401, and the communication station the communication station 2 2403 2 Liu . two content providers and CPC-1 and CPC-2 in this example, the fans will make assumptions:

1、 CMS2410被分配给通信站2401 1, CMS2410 is allocated to the communication station 2401

2、 CMS2411被分配给通信站2403 2, CMS2411 is allocated to the communication station 2403

3、 CPC-1被分配给CMS2410 3, CPC-1 is assigned to the CMS2410

4、 CPC-2被分配给CMS2411 4, CPC-2 is assigned to CMS2411

5、 CPC-1拥有内容C1 (在下文简称为C1) 5, CPC-1 owns Content C1 (hereinafter simply referred to as C1)

6、 CPC-2拥有内容C2 (在下文简称为C2) 6, CPC-2 owns Content C2 (hereinafter simply referred to as C2)

7、 Cl已经被分布到通信站2401和通信站2402 7, Cl has been distributed to the communication station the communication station 2401 and 2402

8、 C2已经被分布到通信站2403和通信站2402。 8, C2 has been distributed to the communication station the communication station 2403 and 2402. 在这个例子中,对于CMS2410给定这些假设,则有: In this example, for the CMS2410 given these assumptions, there are:

1、 内客提供者数据库包括CPC-1 1, the passenger provider database includes CPC-1

2、 内容提供者使用数据库包括在通信站2401、通信站2402和通信站2403上的CPC-1的存储器使用 2, content provider 2401 includes a database using the communication station, CPC-1 memory 2402 and the communication station the communication station 2403 using

3、 文件分布标准数振库包括C1的分布标准 3, document distribution standard library includes vibration number of distribution criteria C1

4、 内容使用和统计数据库包括来自通信站2401、通信站2402和通倌站2403的Cl的使用信息 4, content usage and statistical database includes the use of Cl from the communication station 2401, through communication station 2402 and station 2403 information groom

5、 因为Cl没有分布到通信站2403,将没有来自通信站2403的记 5, since the Cl does not distributed to the communication station 2403, from the communication station will not mind 2403

record

对于CMS2411,刑有: For CMS2411, punishment are:

1、 内容提供者数据库包括CPC-2 1, the content provider database includes CPC-2

2、 内容提供者使用数振库包括在通信站2401、通信站2402和通信站2403上的CPC-2的存储器使用 2, the content provider database using several transducer comprises a memory 2401 in the communication station, the communication station the communication station 2402 and 2403 using the CPC-2

3、 文件分布标准數据库包括C2的分布标准 3, comprises a standard database files are distributed in the distribution criterion C2

4、 内容使用和统计数据库包括来自通信站2401、通信站2402和通信站2403的C2的使用信息 4, includes a content usage and statistical database, the communication station the communication station 2402 and 2403 using the information from the communication station C2 2401

5、 因为C2没有分布到通信站2401,将没有来自通信站2401的记 5, since C2 is not distributed to the communication station 2401, a communication station will not be referred to from 2401

record

对于通信站2401,则有: 2401 for the communication station, there are:

1 、内容提供者数据库包括CPC-1和CPC-2 ,因为内容提供者數据被复制到整个SCDN 1, the content provider database includes CPC-1 and CPC-2, because content provider data is replicated to the entire SCDN

2、内容提供者使用数据库包括在通信站2401上的CPCM和CPC-2 的存储器使用3、 文件分布标准數据库包括C1的分布标准 2, comprises a content provider database and a memory communicatively CPCM station 2401 3 CPC-2 is used, the file distribution criteria database comprises a distribution criterion C1

4、 文件元数据数据库包括C1的文件元数据 4, the file metadata database file including metadata C1

5、 数据块包括C1的内容 5, the data block including the content C1

6、 内容使用和统计数据库包括在通信站2401上的Cl的使用信息对于通信站2402,則有: 6, content usage information and statistical database includes the use of Cl in the communication station the communication station 2401 to 2402, there are:

1 、内容提供者数振库包括CPC-1和CPC-2 1, the content provider database comprises several transducer CPC-1 and CPC-2

2、 内容提供者使用数据库包括在通信站2402上的CPC-1和CPC-2 的存储器使用 2, content providers use the database in the communication station comprises a memory 2402 CPC-1 and CPC-2 is used

3、 文件分布标准数据库包括C1和C2的分布标准 3, includes a distribution file distribution criteria database criteria C1 and C2

4、 文件元数据数据库包括Cl和C2的文件元数据 4, comprises a file metadata database file metadata Cl and C2

5、 数据块包括C1和C2的内容 5, the data blocks comprising a content C1 and C2

6、 内容使用和统计数据库包括在通信站2402上的C1和C2的使 6, and the content usage statistics in the database comprises communication station C1 and C2 2402

用信息 With information

对于通信站2403,则有: 2403 for the communication station, there are:

1 、内容提供者it据库包括CPC-1和CPC-2 1, the content provider database it includes CPC-1 and CPC-2

2、 内容提供者使用数据库包括在通信站2403上的CPOl和CPC-2 2, content providers use the database in the communication station comprises CPOl 2403 and CPC-2

的存储器使用 Memory usage

3、 文件分布标准数据库包括C2的分布标准 3, comprises a standard database files are distributed in the distribution criterion C2

4、 文件元数据数据库包括C2的文件元数据 4, file metadata database includes file metadata C2

5、 数据块包括C2的内容 5, the data block including the content C2

6、 内容使用和统计数据库包括在通信站2403上的C2的使用信息頻带外的信号传输和控制 6, and the content usage statistics database and a control signal transmission comprises an outer band of C2 is used in the information communication station 2403

在本发明的一个实施例中,在每一个分布节点上为应用服务器提供插入模块以使用和定位"頻带外"信号.该插入模块为了访问控制和增值服务可以调用远程控制模块.通过提供这样的一个基础设施,分布网络允许内容提供者在网络的边緣控制它们的内容,并提供动态增值服务. In one embodiment of the present invention, each distribution node in the plug-in module to provide and use targeting "band" signal to the application server. The insertion module for access control and value-added services can invoke remote control modules. By providing such an infrastructure, distribution networks allow content providers at the edge of the network control their content and value-added services to provide dynamic.

访问控制包括由区域进行的鲞別和许可,由用户进行的并行访问的数目,等等.增值服务包括动态的限于局部的和面向目标的广告插入、 动态预览和片尾插入以及更多."頻带外"信号还用作对内容提供者的通知. Access control includes dried sea by the region and do not permit, the number of parallel access by the user, etc. is limited to value-added services include dynamic localized and targeted advertisement insertion oriented, dynamic preview and trailer insertions, and more. "Frequency band "signal also serves as a notification to the content provider.

Claims (20)

1. 一种用于将大有效负载文件的内容分布到一个网络节点中的多个存储设备的方法,包括:获得代表一个大有效负载文件的内容的多个块文件,以便存储在一个网络节点中的多个存储设备中;将所述多个块文件与所述多个存储设备相联系;基于所述联系将所述多个块文件存储在所述多个存储设备中;以及利用所述联系创建一个虚拟文件,以便展示给一个请求所述文件内容的客户机,所述虚拟文件向所述客户机提供一个所述文件内容在所述网络节点中是连续的假象,其中,创建所述虚拟文件包括:从文件元数据中获得有关所述大有效负载文件的信息;和使用所述信息来构造所述大有效负载文件的一个虚拟表示。 1. A method for the large payload file content distribution method of a network node to a plurality of storage devices, comprising: obtaining a plurality of blocks representative of a large payload file content file, so that a network node stores a plurality of storage devices; the plurality of blocks of the file associated with said plurality of storage devices; link based on the plurality of storage devices to the plurality of blocks in the file storage; and using the Contact create a virtual file a request for presentation to the client of the contents of the file, the file provides a virtual file content is the illusion of continuous node in the network to the client, which created the the virtual file comprising: obtaining information about the large payload file from the file metadata; and the use of the information to construct a virtual representation of the large payload files.
2. 如权利要求1所述的方法,其中,所述获得多个块文件包括:从一个用户获得所述大有效负载文件;评价所述大有效负载文件,以定位一个具有实质性内容的部分;通过确定所述实质性内容是否位于所述大有效负栽文件的开始端来确定所述来自所述大有效负栽文件的内容是否包括线性特性;如果所述内容包括所述线性特性,则产生一个磁道文件;如果所述内容不包括所述线性特性,则产生多个磁道文件;确定至少一个期望的块大小;以及如果所述内容包括线性特性,划分所述磁道文件以产生多个块文件,其中,所述块文件中的每一个在大小上对应于所述至少一个期望的块大小;以及如果所述内容不包括线性特性,划分所述多个磁道文件中的每一个以产生多个块文件,其中,所述块文件中的每一个在大小上对应于所述至少一个期望的块大小。 Evaluating the portion of the large payload file, to locate a substantive content; obtaining a large payload file from a user: The method according to claim 1, wherein said file comprises a plurality of blocks obtained ; determining the content from said large payload file plant comprises a linear characteristic by determining whether the substance is located at the starting end of the large payload file plant; if said content comprises said linear characteristics, the generating a track file; if the content does not include the linear characteristic, a plurality of tracks are generated file; determining at least one desired block size; and if said content comprises linear characteristics dividing said track file to generate a plurality of blocks document, wherein the size of each block in size corresponding to the at least one desired block file; and if the content does not comprise linear characteristics dividing said plurality of tracks for each file to generate a plurality file blocks, wherein the blocks of the file in each of a size corresponding to the at least one desired block size.
3. 如权利要求2所述的方法,其中,通过选择一个块大小以使所述网络的最大传输单位内的空间使用率最大来确定所述期望的块大小。 The method according to claim 2, wherein, by choosing a block size so that the space within the maximum transmission unit of the network to determine a maximum utilization of the desired block size.
4.如权利要求2所述的方法,其中,所述网络具有多个分布节点,并且通过选择一个块大小以使在所述网络的所述分布节点上的拥塞最小来确定所述期望的块大小。 4. The method according to claim 2, wherein said network having a plurality of distributed nodes, and by selecting a block size so that congestion on the distribution node of the network to determine the desired minimum block size.
5.如权利要求1所述的方法,其中,所述联系所述多个块文件包括在所述多个存储设备之间分布所述多个块文件,以使得所述多个存储设备在输/J输出操作期间被负栽均衡. 5. The method according to claim 1, wherein said contact of said plurality of blocks of said file comprises a plurality of blocks distributed between said plurality of file storage devices, so that the plurality of storage devices in the output during / J plant is negatively balanced output operation.
6. 如权利要求1所述的方法,进一步包括将所述联系存储在所述多个存储设备中的文件元数据中,所述联系包括用于从所述块文件重建所述大有效负栽文件的信息. 6. The method according to claim 1, further comprising a contact file metadata stored in the plurality of storage devices, said contact comprising a block from said large payload file rebuild the plant information file.
7. 如权利要求1所述的方法,其中,所述网络包括多个节点,并且所述多个块文件是从所述网络的其他节点中获得的。 7. The method according to claim 1, wherein said network comprises a plurality of nodes, and the plurality of blocks is obtained from the files of the other nodes in the network.
8. —种用于将大有效负栽文件的内容分布到一个网络节点中的多个存«备的装置,包括:一个网络节点的笫一服务器簇中的一个或更多第一服务器,获得代表一个大有效负栽文件的内容的多个块文件,用于存储在所述网络节点中的多个存储设备中,所述一个或更多第一服务器将所述多个块文件与所述多个存储设备相联系,并基于所述联系将所述多个块文件存储在所述多个^^:备中;以及所述网络节点的第二服务器簇中的一个或更多第二服务器,利用所述联系创建一个虚拟文件,包括从文件元数据中获得有关所述大有效负栽文件的信息,使用所述信息来构造所述大有效负栽文件的一个虚拟表示,以氣艮示给一个请求所述文件内容的客户机,所述虚拟文件向所述客户Wi供一个所述文件内容在所述网络节点中是连续的假象. 8. - kind of content for a large payload file is distributed to a plant means a plurality of network nodes stored «apparatus, comprising: a server cluster Zi one network node to one or more of the first server, to obtain representing a large payload file contents of a plurality of blocks planted files stored in a plurality of storage devices in the network node, the one or more of the plurality of blocks of the first file server and said associated plurality of storage devices, and the plurality of contact based on the file stored in the plurality of blocks ^^: Preparation; and a second cluster server of the network node or more second server using said contact creates a virtual file, comprising obtaining information about the plant a large payload file from file metadata, using the information to construct a large payload file is a virtual representation of the plant, in order to illustrate the gas gen requesting the file content to a client, the virtual file for a Wi content of the file to the client the illusion of a continuous network node.
9. 如权利要求8所述的装置,其中,所«得多个块文件包括:笫三服务器从一个用户获得所述大有效负载文件,所述第三服务器评价所述大有效负栽文件,以定位一个具有实质性内容的部分,并通过确定所述实质性内容是否位于所述大有效负载文件的开始端来确定所述来自所述大有效负栽文件的内容是否包括线性特性,如果所述内容包括所述线性特性,则所述第三服务器产生一个磁道文件并将所述磁道文件发送到所述一个或更多第一服务器,而如果所述内容不包括所述线性特性,则所述第三服务器产生多个磁道文件并将所述多个磁道文件发送到所述一个或更多笫一服务器;以及所述一个或更多第一服务器以一个或更多磁道文件的形式接收所述大有效负栽文件,所述一个或更多第一服务器确定用于将所述一个或更多磁道文件划分成多个块文件的至少一 9. The apparatus according to claim 8, wherein the «blocks have a plurality of files comprising: a server obtains the sleeping mat three large payload file from a user, the third server evaluating the large payload file plant, to locate a portion having substantial contents, and determining from the contents of the large payload file plant comprises a linear characteristic by determining whether the substance is located at the starting end of the large payload file, if transmitting said content comprises said linear characteristics, then the server generates a third track of the track file and the file to the one or more first servers, and if the content does not include the linear characteristic, then the said third plurality of tracks file server generates and sends the file to the plurality of tracks one or more Zi server; and said one or more first server received in the form of one or more of the document track said large payload file planted, the one or more first servers for determining the one or more tracks is divided into a plurality of files at least a block of the file 期望的块大小,其中,所迷块文件之中的每一个块文件在大小上对应于所述至少一个期望的块大小。 Desired block size, wherein each block of the file from among the files on the fan block size corresponding to the at least one desired block size.
10. 如权利要求8所述的装置,其中,所述联系所述多个块文件包括在所述多个存储设备之间分布所述多个块文件,以使得所述多个存储设备在输A/输出操作期间被负载均衡. 10. The apparatus according to claim 8, wherein said file comprises contacting the plurality of blocks of said plurality of blocks distributed between said plurality of file storage devices, so that the plurality of storage devices in the output during the A / output operation is load balancing.
11. 如权利要求8所述的装置,进一步包括所述一个或更多笫一服务器将所述联系存储在所述多个存储设备中的文件元数据中,所述联系包括用于从所述块文件中重建所述大有效负栽文件的信息。 Comprising from said apparatus as recited in claim 8, further comprising one or more Zi, a contact server, the file metadata stored in the plurality of storage devices, the contact reconstructing the large block file payload information file plant.
12. 如权利要求11所述的装置,其中,所述创建一个虚拟文件包括:从所述文件元数据中获得有关所述大有效负载文件的信息;以及使用所述信息来构造所述大有效负栽文件的一个虚拟表示。 The large effective to construct and use of the information; get information about the large payload file from the file metadata: 12. The apparatus as claimed in claim 11, wherein said creating a virtual file comprises negative planted a virtual representation of the file.
13. —种用于从不连续分段的块文件中产生一个大有效负载文件的方法,包括:在一个网络节点上接收对一个大有效负栽文件的请求;获得有关汇编所述大有效负栽文件的信息;从与所述网络节点相联系的多个存储设备中获得所述大有效负载文件的多个块文件;使用所述汇编信息从所述多个块文件中创建所述大有效负栽文件的一个虚拟表示,其中,所迷虚拟表示创建步骤包括根据所述汇编信息创建所述多个块文件的一个虚拟汇编和所述多个块文件的表示的线性叠加;以及响应于所述请求,发送所述大有效负栽文件. 13. - Method a large payload file block for the file is generated species from discontinuous segments, comprising: receiving at a network node requests a large payload file plant; obtaining the compilation of the large payload plant information file; obtaining a plurality of blocks of the large payload file from a file associated with the plurality of storage devices in the network node; using the compiled information to create the large effective from the plurality of block file a virtual file planted negative representation, wherein the fan comprises the step of creating the virtual representation represents a linear superposition of a plurality of blocks of said virtual file compilation and a plurality of blocks of a file created based on the compiled information; and in response to the said request, transmitting said large payload file plant.
14. 如权利要求13所述的方法,其中,所述汇编信息被包含在存储在所述节点的所述多个存储设备中的一个文件元数据数据库中, 14. The method according to claim 13, wherein said compiling information contained in one file is stored in the storage device of said plurality of nodes in the metadata database,
15. 如权利要求13所述的方法,其中,所迷块文件以不连续形式存储在所述多个存储设备中。 15. The method according to claim 13, wherein the discontinuous fans block file stored in said plurality of storage devices.
16. —种用于以在多个网络节点之间分布的内容的各部分来满足对所述内容的一个客户机请求的方法,包括:在第一节点上从一个客户机接收对内容的第一请求,其中,所述第一节点包括一个网络的多个节点中的一个节点,所述第一节点具有所述内容的一部分,其中,所述内容在所述多个节点中的一个或更多节点之间被分布在块文件的子集中;使用在所述第一节点中可得到的所述内容的一个文件元数据中的数据来准备所述内容的一个虚拟图像,所述文件元数据具有有关重新汇编所述文件的信息,所述虚拟图像向所ii^户机提供一个所述内容在所述第一节点上可完全得到的假象,其中,所述虚拟困像创建包括从所述文件元数据中获得有关所述大有效负载文件的信息和使用所述信息来构造所述大有效负栽文件的一个虚拟图像;从在所述第一节点上可得到的 16. - for species in each portion between a plurality of network nodes distributed content to satisfy client to a method of the content request, comprising: receiving first content from a client at the first node a request, wherein the first node comprises a plurality of nodes in a network, said first portion having a point of the content, wherein the content of said one or more of the plurality of nodes sub-block is distributed among multiple files centralized node; a virtual image data of a document using the content metadata available at the first node in preparing the content, the metadata file having the information about the reassembled document, providing the illusion of a virtual image of the content on the first node can be fully obtained in ii ^ to the client, wherein the virtual image created comprises from the storm metadata file information and using the information obtained about the large payload file to construct a virtual image of the large payload file plant; available from the first node in the 所述内容的所述部分开始,开始传输所述内^所i^户机;在所述传输正在进行的同时,获得所述内容的非驻留部分,所述获得非驻留部分包括:从第一节点发送一个第二请求,以确定所述多个节点之中具有所述内容的所述非驻留部分的节点,其中,所述多个节点中的一个或更多节点可以具有所述内容的相同部分;从具有所述内容的每个响应节点接收一个响应,所述响应指定所述内容的部分和所述响应节点的性能特性;从所述响应中确定所述响应节点之中的哪些节点是下载所述内容的所述非驻留部分的一个期望集合的节点;以及从所述期望集合的节点中下栽所述内容的所述非驻留部分,并将所述非驻留部i^存储在所述请求节点上。 The portion of the content is started, the start of the transmission of the I ^ ^ client; while the transfer is in progress, to obtain non-resident portion of the content, the non-resident portion of the obtaining comprises: from the first node sends a second request to determine the non-resident node portions among the plurality of nodes with the content, wherein the plurality of nodes, one or more nodes may have the the same portion of the content; receiving a response from each of the responding node having content, the response content of the designated portion of the response and performance characteristics of the node; in response to determining that the response from the node which node is the node to download the content to a desired non-resident portion of the set; and the non-resident portion of the content from the tilted downward and crashed desired set of nodes, and the non-resident i ^ portion is stored in the requesting node.
17. —种用于从不连续分段的块文件中产生一个大有效负载文件的装置,包括:一个具有多个节点的网络;在所述多个节点之中的每一个节点中的多个存储设备;以及所述网络的一个节点的笫一服务器簇中的第一服务器,接收对一个大有效负栽文件的内容的请求,所述第一服务器簇具有一个或更多笫一服务器,所述笫一服务器获得有关汇编所述大有效负载文件的信息,所述第一服务器从所述节点的所述多个存储设备中获得所述大有效负载文件的多个块文件,所述第一服务器使用所述汇编信息从所述多个块文件中创建所述大有效负栽文件的一个虚拟表示,包括根据所述汇编信息创建所述多个块文件的一个虚拟汇编和所述多个块文件的表示的线性叠加;并响应于所述请求来发送所述大有效负载文件。 - 17. A block file types for the device from discrete segments of a large payload file comprises generating: a network having a plurality of nodes; among a plurality of nodes each of the plurality of nodes in a storage device; and the undertaking of a network node of a server cluster first server, receiving a request of a large payload file contents plant, the first server having a cluster of servers or more Zi, the Zi said server obtains information about a large payload to the compiled file, the first file server to obtain a plurality of blocks of the large payload file from the storage device of said plurality of nodes, said first the server uses the information to create a virtual compilation of the large payload file from a plant of the plurality of blocks represent files, including the compiled information according to the plurality of blocks to create a virtual file and the compilation of a plurality of blocks It represents a linear superposition of the document; and in response to the request to transmit the large payload files.
18. 如权利要求17所述的装置,其中,所述汇编信息被包含在存储在所述节点的所述多个存储设备中的一个文件元数据数据库中。 18. The apparatus according to claim 17, wherein said compiling information contained in one file is stored in the storage device of said plurality of nodes in the metadata database.
19. 如权利要求17所述的装置,其中,所述块文件以不连续形式存储在所迷多个存储设备中。 19. The apparatus according to claim 17, wherein said block storage file in the form of a discontinuous fans of the plurality of storage devices.
20.如权利要求17所述的装置,其中,所述虚拟表示包括所述多个块文件的一个虚拟汇编,包括根据所述汇编信息的所述多个块文件的表示的线性叠加. 20. The apparatus according to claim 17, wherein the virtual representation comprising a plurality of blocks of said virtual file compilation, including linear superposition of the compilation of information representing a plurality of blocks of a file.
CNB018208142A 2000-10-26 2001-10-26 Method and apparatus for large payload distribution in network CN100525288C (en)

Priority Applications (16)

Application Number Priority Date Filing Date Title
US26628600P true 2000-10-26 2000-10-26
US60/266,286 2000-10-26
US09/681,644 US6970939B2 (en) 2000-10-26 2001-05-15 Method and apparatus for large payload distribution in a network
US09/681,644 2001-05-15
US09/681,665 2001-05-17
US09/681,665 US7165095B2 (en) 2000-10-26 2001-05-17 Method and apparatus for distributing large payload file to a plurality of storage devices in a network
US09/681,669 US7181523B2 (en) 2000-10-26 2001-05-18 Method and apparatus for managing a plurality of servers in a content delivery network
US09/681,671 2001-05-18
US09/681,669 2001-05-18
US09/681,671 US6857012B2 (en) 2000-10-26 2001-05-18 Method and apparatus for initializing a new node in a network
US09/681,672 2001-05-18
US09/681,673 US7058014B2 (en) 2000-10-26 2001-05-18 Method and apparatus for generating a large payload file
US09/681,668 US7177270B2 (en) 2000-10-26 2001-05-18 Method and apparatus for minimizing network congestion during large payload delivery
US09/681,668 2001-05-18
US09/681,672 US7047287B2 (en) 2000-10-26 2001-05-18 Method and apparatus for automatically adapting a node in a network
US09/681,673 2001-05-18

Publications (2)

Publication Number Publication Date
CN1481635A CN1481635A (en) 2004-03-10
CN100525288C true CN100525288C (en) 2009-08-05

Family

ID=27575312

Family Applications (1)

Application Number Title Priority Date Filing Date
CNB018208142A CN100525288C (en) 2000-10-26 2001-10-26 Method and apparatus for large payload distribution in network

Country Status (4)

Country Link
EP (1) EP1368948A2 (en)
CN (1) CN100525288C (en)
AU (1) AU2444802A (en)
WO (1) WO2002035799A2 (en)

Families Citing this family (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
AT381191T (en) 2000-10-26 2007-12-15 Prismedia Networks Inc A method and system for managing distributed contents and corresponding metadata-
US7171469B2 (en) 2002-09-16 2007-01-30 Network Appliance, Inc. Apparatus and method for storing data in a proxy cache in a network
US7284030B2 (en) * 2002-09-16 2007-10-16 Network Appliance, Inc. Apparatus and method for processing data in a network
US7552223B1 (en) 2002-09-16 2009-06-23 Netapp, Inc. Apparatus and method for data consistency in a proxy cache
KR100585668B1 (en) 2004-03-22 2006-06-07 엘지전자 주식회사 Multimedia compiling method and apparatus using virtual media file system and multimedia file structure
US7756051B2 (en) 2004-07-02 2010-07-13 Microsoft Corporation Content distribution using network coding
CN100433619C (en) 2004-07-23 2008-11-12 腾讯科技(深圳)有限公司 Data download method and system
CN100411341C (en) 2005-08-10 2008-08-13 华为技术有限公司 Parallel downloading method and terminal
EP1929738B1 (en) * 2005-09-08 2013-01-16 Nortel Networks Limited Load balancing for an air interface protocol architecture with a plurality of heterogenous physical layer modes
US20070168405A1 (en) 2006-01-17 2007-07-19 Ori Pomerantz Self-optimizing network attached storage for multiple geographic locations
CN101064729B (en) 2006-04-27 2010-06-09 中国电信股份有限公司 System and method for realizing FTP download service through CDN network
WO2007143882A1 (en) * 2006-06-09 2007-12-21 Fast Dragon International Holdings Limited N^n data management, access, storage, transfer, exchange and retrieval system (data master)
CN101184104B (en) 2007-12-21 2010-12-15 腾讯科技(深圳)有限公司 Distributed memory system and method
JP4557028B2 (en) 2008-03-19 2010-10-06 ソニー株式会社 The information processing apparatus, information processing method, client device, information processing system
CA2673554C (en) * 2009-07-21 2017-01-03 Ibm Canada Limited - Ibm Canada Limitee Web distributed storage system
US8489654B2 (en) 2009-08-28 2013-07-16 Beijing Innovation Works Technology Company Limited Method and system for forming a virtual file system at a computing device
US8612622B2 (en) * 2009-10-02 2013-12-17 Limelight Networks, Inc. Real-time message queuing for a processing ring
CN102486739B (en) * 2009-11-30 2015-03-25 国际商业机器公司 Method and system for distributing data in high-performance computer cluster
US9628579B2 (en) 2010-05-13 2017-04-18 Futurewei Technologies, Inc. System, apparatus for content delivery for internet traffic and methods thereof
US8463788B2 (en) * 2010-09-03 2013-06-11 Marvell World Trade Ltd. Balancing caching load in a peer-to-peer based network file system
JP5743469B2 (en) * 2010-09-22 2015-07-01 キヤノン株式会社 The information processing apparatus, a control method, and control program
US8612550B2 (en) 2011-02-07 2013-12-17 Microsoft Corporation Proxy-based cache content distribution and affinity
US20120314575A1 (en) * 2011-04-15 2012-12-13 Deutsche Telekom Ag Network traffic engineering
CN103563327B (en) * 2011-06-08 2017-07-11 皇家Kpn公司 The method of locating and retrieving segmented content and client
WO2013096758A1 (en) 2011-12-21 2013-06-27 Akamai Technologies, Inc. Security policy editor
JPWO2013146808A1 (en) * 2012-03-28 2015-12-14 日本電気株式会社 Computer system, and a communication path change method
EP2660723A1 (en) * 2012-05-03 2013-11-06 Thomson Licensing Method of data storing and maintenance in a distributed data storage system and corresponding device
US9246998B2 (en) * 2012-10-16 2016-01-26 Microsoft Technology Licensing, Llc Load balancer bypass
CN106850817A (en) * 2012-12-10 2017-06-13 北京奇虎科技有限公司 Download management device, method and data download system
CN103152396A (en) * 2013-02-05 2013-06-12 华南师范大学 Data placement method and device applied to content distribution network system
CN103778212B (en) * 2014-01-16 2017-04-05 国网山东省电力公司青岛供电公司 Massive parallel data processing method based on the data node
CN103929491B (en) * 2014-04-28 2016-02-03 努比亚技术有限公司 Method and system for data synchronization and the mobile terminal
CN105634992B (en) * 2015-12-29 2019-01-11 网宿科技股份有限公司 CDN platform adaptive band width control method and system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5727002A (en) 1995-01-19 1998-03-10 Starburst Communications Corporation Methods for transmitting data
JP2000148711A (en) 1998-11-10 2000-05-30 Victor Co Of Japan Ltd Dynamic image server system
US6081840A (en) 1997-10-14 2000-06-27 Zhao; Yan Two-level content distribution system

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
IL115263A (en) * 1995-09-12 1999-04-11 Vocaltec Ltd System and method for distributing multi-media presentations in a computer network
US6038601A (en) * 1997-07-21 2000-03-14 Tibco, Inc. Method and apparatus for storing and delivering documents on the internet
US6105029A (en) * 1997-09-17 2000-08-15 International Business Machines Corporation Retrieving network files through parallel channels
US20010052015A1 (en) * 1998-06-24 2001-12-13 Chueng-Hsien Lin Push-pull sevices for the internet

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5727002A (en) 1995-01-19 1998-03-10 Starburst Communications Corporation Methods for transmitting data
US6081840A (en) 1997-10-14 2000-06-27 Zhao; Yan Two-level content distribution system
JP2000148711A (en) 1998-11-10 2000-05-30 Victor Co Of Japan Ltd Dynamic image server system

Also Published As

Publication number Publication date
WO2002035799A3 (en) 2003-09-25
AU2444802A (en) 2002-05-06
WO2002035799A2 (en) 2002-05-02
EP1368948A2 (en) 2003-12-10
CN1481635A (en) 2004-03-10

Similar Documents

Publication Publication Date Title
Acharya et al. Middleman: A video caching proxy server
US7546364B2 (en) Replication of remote copy data for internet protocol (IP) transmission
US9930099B2 (en) Hybrid segment-oriented file server and WAN accelerator
KR101381014B1 (en) Distributed replica storage system with web services interface
CN102143215B (en) Network-based PB level cloud storage system and processing method thereof
US7324555B1 (en) Streaming while fetching broadband video objects using heterogeneous and dynamic optimized segmentation size
US8234327B2 (en) System and method for bandwidth optimization in a network storage environment
EP2409240B1 (en) Variable rate media delivery system
EP2411918B1 (en) Virtualized data storage system architecture
Borthakur The hadoop distributed file system: Architecture and design
US6182111B1 (en) Method and system for managing distributed data
US9992279B2 (en) Managed object replication and delivery
US6449731B1 (en) Self-healing computer system storage
US8190741B2 (en) Customizing a namespace in a decentralized storage environment
EP2266043B1 (en) Cache optimzation
US7822711B1 (en) Conflict resolution for a distributed file sharing system
EP1540498B1 (en) Apparatus and method for proxy cache
CN101668046B (en) Resource caching method, device and system thereof
US8234372B2 (en) Writing a file to a cloud storage solution
US8447842B2 (en) Scalable, high performance and highly available distributed storage system for internet content
US7761900B2 (en) Distribution of content and advertisement
US6167438A (en) Method and system for distributed caching, prefetching and replication
EP1330907B1 (en) Method and apparatus for real-time parallel delivery of segments of a large payload file
US9135269B2 (en) System and method of implementing an object storage infrastructure for cloud-based services
EP2169909B1 (en) System and method to maintain coherence of cache contents in a multi-tier software system aimed at interfacing large databases

Legal Events

Date Code Title Description
C06 Publication
C10 Entry into substantive examination
C14 Grant of patent or utility model
CF01