CN102073593B - Method and device for transmitting data block in memory, method and device for receiving data block in memory and transmission system - Google Patents

Method and device for transmitting data block in memory, method and device for receiving data block in memory and transmission system Download PDF

Info

Publication number
CN102073593B
CN102073593B CN 201010546222 CN201010546222A CN102073593B CN 102073593 B CN102073593 B CN 102073593B CN 201010546222 CN201010546222 CN 201010546222 CN 201010546222 A CN201010546222 A CN 201010546222A CN 102073593 B CN102073593 B CN 102073593B
Authority
CN
China
Prior art keywords
data
memory
packet
block
receiving
Prior art date
Application number
CN 201010546222
Other languages
Chinese (zh)
Other versions
CN102073593A (en
Inventor
李海东
施培任
Original Assignee
浪潮(北京)电子信息产业有限公司
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
Application filed by 浪潮(北京)电子信息产业有限公司 filed Critical 浪潮(北京)电子信息产业有限公司
Priority to CN 201010546222 priority Critical patent/CN102073593B/en
Publication of CN102073593A publication Critical patent/CN102073593A/en
Application granted granted Critical
Publication of CN102073593B publication Critical patent/CN102073593B/en

Links

Abstract

本发明提供一种内存中数据块的发送、接收方法和装置及传输系统,涉及通信领域;能够缩短数据块的传输时间。 The present invention provides a method of transmitting data blocks in memory, receiving method and apparatus and a transmission system, the field of communication; possible to shorten the transmission time of a data block. 所述发送方法,包括:当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间;为本地用于缓冲所述数据包中全部数据块的内存空间进行注册;当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中;通过无限带宽链路向所述接收端发送所述工作请求。 The transmission method, comprising: when detecting the data block of the data packet need unlimited bandwidth link transmission, the receiving side notifies register for storing said data packet in memory all the data blocks; local buffering for all memory registers of said data blocks in the data packet; and when the detected local receiving end to complete the registration memory space, filling the entire data block of the data packet to a work request; chain by Infiniband transmitting said work request path to the receiving end. 本发明提供的技术方案可应用于信息传输。 The present invention provides a technical solution can be applied to information transmission.

Description

内存中数据块的发送、接收方法和装置及传输系统 Transmitting the data block in memory, receiving method and apparatus and a transmission system

技术领域 FIELD

[0001] 本发明涉及计算机应用领域,尤其涉及一种在内核运行态下内存中数据块的发送、接收方法和装置及传输系统。 [0001] The present invention relates to the field of computer applications, and particularly to a transmitting, receiving apparatus and transmission method and system memory data blocks in kernel operating mode.

背景技术 Background technique

[0002] Infiniband(无限带宽)是一种高速互联技术,用于作系统内部互联,替代总线的使用。 [0002] Infiniband (unlimited bandwidth) is a high speed interconnect technology, used as an internal interconnection system, instead of using the bus. 其中该技术还拥有RDMA(Remote Direct Memory Access,远程直接内存访问)的特性。 The technology also has properties which RDMA (Remote Direct Memory Access, Remote Direct Memory Access) is.

[0003] 在实现本发明过程中,发明人发现现有技术存在如下问题: [0003] In implementing the present invention, the inventors have found that the prior art has the following problems:

[0004] 现有技术中使用Infiniband链路传输数据块时,采用RDMA进行内存管理,要求每次传输的数据块必须是页对齐的数据块,而采用该机制时,在数据包中存在首数据块相对于页右不对齐,尾数据块相对于页左不对齐,中间数据块相对于页不对齐这三种中的任意一种情况时,就需要将整个数据包分多次发送,增加了消息往返交互的次数,从而增加了数据包中全部数据块的传输时间,降低了数据包的传输效率。 When [0004] the prior art using Infiniband link transmission data blocks, using RDMA memory management, requires each block transfer page data block must be aligned, and when using this mechanism, in the presence of a first data packet block with respect to misalignment of the right page, the left end of the data block is not aligned with respect to the page, with respect to the intermediate block is not aligned with any of the pages of these three in the case, it is necessary to transmit the whole packet multiple times, increased message to and from the number of interactions, thus increasing the transmission time all the packets of data blocks, reduces the transmission efficiency of data packets.

发明内容 SUMMARY

[0005] 本发明提供一种在内核运行态下内存中数据块的发送、接收方法和装置及传输系统,缩短数据块的传输时间。 [0005] The present invention provides a transmitting apparatus and a receiving method and transmission system memory data blocks in kernel operating mode, reduce transmission time of the data block.

[0006] 为解决上述技术问题,本发明提供了如下技术方案: [0006] In order to solve the above technical problem, the present invention provides the following technical solutions:

[0007] 一种在内核运行态下内存中数据块的发送方法,包括: [0007] A method of transmitting data blocks in the memory kernel operating mode, comprising:

[0008] 当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间; [0008] When detecting the data block packet need unlimited bandwidth link transmission, the receiving side notifies register for storing said data packet in memory all the data blocks;

[0009] 为本地用于缓冲所述数据包中全部数据块的内存空间进行注册; [0009] The local memory space for buffering said data packet in all the data block to be registered;

[0010] 当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中; [0010] Upon detecting the local and the receiving end to complete the registration memory, the entire data block of the data packet is filled into a work request;

[0011] 通过无限带宽链路向所述接收端发送所述工作请求。 [0011] The work request sent to the receiving end by an infinite bandwidth link.

[0012] 优选的,在所述通知接收端注册用于存储所述数据包中全部数据块的内存空间时还包括: [0012] Preferably, the notification receiving side register for storing said data packet in memory all the data blocks it further includes:

[0013] 通知所述接收端所述数据包中全部数据块的格式信息,其中所述格式信息为本地和所述接收端预先为每个格式信息的描述信息设置的字符数小于该格式信息的描述信息字符数的标识。 [0013] The format of the notification information packet in said block received all of the data terminal, wherein said format information is less than the local format information for the receiving end and the number of characters set in advance for each of the description information format information describe the identification number of message character.

[0014] 优选的,所述为本地用于缓冲所述数据包中全部数据块的内存空间进行注册,包括: [0014] Preferably, the local buffering for the data packet in memory to register the entire data block, comprising:

[0015] 在未检测到接收端对所述数据包中数据块的内存注册完成前,为本地用于缓冲所述数据包中全部数据块的内存空间进行注册。 [0015] In the receiving end is not detected prior to the registration of memory data in the data packet block is completed, a local buffer memory of the data packet to register all the data blocks.

[0016] 一种在内核运行态下内存中数据块的接收方法,包括:[0017] 获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知; [0016] A method of receiving at the kernel operating mode of the data memory block, comprising: [0017] Get register for storing the data packets in an InfiniBand link transmission data of all memory blocks comprising a notification;

[0018] 为本地用于存储所述数据包中全部数据块的内存空间进行注册; [0018] registered as a local memory space for storing all the data in the data packet block;

[0019] 在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 [0019] After receiving the data packet comprises a work request by all the data blocks unlimited bandwidth link, the data block of the data packet stored in the memory space of the local registration obtained.

[0020] 一种在内核运行态下内存中数据块的发送装置,包括: [0020] A transmission apparatus operating in kernel mode memory block, comprising:

[0021] 通知模块,用于当检测到有数据包的数据块通过无限带宽链路需要传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间; When the [0021] notification module, configured to, when the data block is detected by the data packet transmission unlimited bandwidth link needs to notify the receiving side register for storing the data packet in memory all the data blocks;

[0022] 注册模块,与所述通知模块相连,用于为本地用于缓冲所述数据包中全部数据块的内存空间进行注册; [0022] The registration module, the notification module is connected, for the data packet to buffer memory all the data blocks to local register;

·[0023] 填充模块,与所述注册模块相连,用于当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中; * [0023] filling module, coupled with the registration module, for detecting when the receiving end and the local memory space completed registration, all of the data packets to the data block to fill a work request;

[0024] 发送模块,与所述填充模块相连,用于通过无限带宽链路向所述接收端发送所述工作请求。 [0024] The transmitting module, connected to the filling module, for transmitting said work request to the receiving end by an infinite bandwidth link.

[0025] 优选的,所述通知模块,还用于通知所述接收端所述数据包中全部数据块的格式信息,其中所述格式信息为本地和所述接收端预先为每个格式信息的描述信息设置的字符数小于该格式信息的描述信息字符数的标识。 [0025] Preferably, the notification module is further configured to format information in the data packet all of the data blocks to be notified to the receiving terminal, wherein the format information for the local and the receiving end in advance for each format of the information identifying the number of characters is smaller than the description set format information is information of the character described.

[0026] 优选的,所述注册模块,用于在未检测到接收端对所述数据包中数据块的内存注册完成前,为本地用于缓冲所述数据包中全部数据块的内存空间进行注册。 [0026] Preferably, the registration module is configured to prior to the reception side does not detect the completion of registration memory data blocks in the data packet, said local packet for buffering all of the data memory block is registered.

[0027] 一种在内核运行态下内存中数据块的接收装置,包括: [0027] A receiving apparatus operating in kernel mode memory block, comprising:

[0028] 获取模块,用于获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知; [0028] obtaining module, configured to obtain notification includes register for storing the data packets in an InfiniBand link transmission of all the memory space of the data block;

[0029] 注册模块,与所述获取模块相连,用于为本地用于存储所述数据包中全部数据块的内存空间进行注册; [0029] The registration module, connected to the obtaining module is configured for storing the local data packet memory to register all the data blocks;

[0030] 存储模块,与所述注册模块相连,用于在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 [0030] The storage module, connected with the registration module, for after receiving the data packet comprises a work request by all the data blocks unlimited bandwidth link, storing the data block of the data packet to a local register the resulting memory space.

[0031] 一种在内核运行态下内存中数据块的接收装置,包括发送装置和接收装置,其中: [0031] A receiving apparatus operating in kernel mode memory block, comprising a transmitting apparatus and a receiving apparatus, wherein:

[0032] 所述发送装置,用于当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间;为本地用于缓冲所述数据包中全部数据块的内存空间进行注册;当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中;通过无限带宽链路向所述接收端发送所述工作请求。 [0032] The transmitting means for detecting when the data block packet requires infinite bandwidth link transmission, the receiving side notifies register for storing the data packet all of the memory space of the data block; with local buffering the data packet to all the memory space of the data block register; Upon detecting the local and the receiving end to complete the registration memory space, filling the entire data block of the data packet to a work request; by Infiniband said work request link transmission to the receiving end.

[0033] 所述接收装置,用于获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知;为本地用于存储所述数据包中全部数据块的内存空间进行注册;在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 [0033] The receiving means for obtaining by comprising registration notification packet for storing unlimited bandwidth link transmission of all the data memory block; local memory for storing the entire data block of the data packet space register; after receiving the data packet comprises a work request by all the data blocks unlimited bandwidth link, storing the data block packet into the memory space obtained locally registered.

[0034] 本发明提供的实施例,发送端将所述数据包的全部数据块通过一个工作请求发送到所述数据包的接收端,与现有技术首先判断是否有页不对齐数据,只传输页对齐数据相t匕,减少了发送同一数据包中全部数据块的时间;另一方面,由于接收端一次性能够接收到数据包中全部数据块,与现有技术中在数据包中有页不对齐的数据块时需要多次接收相t匕,减少了接收同一数据包中全部数据块的次数,缩短了数据块的接收时间,从而达到缩短数据包中全部数据块的传输时间的目的,提高了数据包的传输效率。 [0034] Example embodiments, the transmit end all data blocks of the data packet transmitted to the receiving end of the request packet through a working, prior art to the present invention there is provided a first determines whether the page data is not aligned, transfer only t phase data page-aligned dagger, reducing the time to transmit all data blocks in the same data packet; on the other hand, since the receiving end can receive a one-time all the data packets in the data block, the prior art there are pages in the data packet require multiple receives a data block is not aligned with t dagger, reducing the number of all the received data blocks in the same data packet, reducing the reception time data block, so as to achieve the purpose of shortening the transmission time of all data blocks in the data packet, improved efficiency of data packet transmission.

[0035] 进一步的,在发送端通知接收端进行内存注册后立即进行本地的内存注册,使得发送端与接收端的内存注册几乎同时进行,缩短了整体的内存注册时间,也达到了缩短整个数据包中数据块的传输时间。 After [0035] Further, the transmitting side notifies the receiving end memory register immediate local memory register, so that the transmission and the receiver memory register almost simultaneously, shortening overall memory registration time has reached shorten the entire packet transmission time of a data block.

附图说明 BRIEF DESCRIPTION

[0036] 图I为本发明提供的在内核运行态下内存中数据块的发送方法实施例的流程示意图; [0036] FIG I in the flow diagram of an embodiment of the kernel operating mode memory block transmission method of the present invention provides;

[0037] 图2为本发明提供的在内核运行态下内存中数据块的接收方法实施例的流程示意图; [0037] FIG 2 flow of the embodiment method for receiving operating mode kernel memory block schematic diagram of the present invention to provide;

[0038] 图3为本发明提供的在内核运行态下内存中数据块的传输方法应用实例的流程示意图; [0038] FIG. 3 Process runs in kernel mode memory transfer method in a data block schematic application example of the present invention provides;

[0039] 图4为本发明提供的在内核运行态下内存中数据块的发送装置实施例的结构示意图; [0039] FIG. 4 configuration diagram of an embodiment in kernel memory apparatus operating mode data transmission provided by the present invention, the block;

[0040]图5为本发明提供的在内核运行态下内存中数据块的接收装置实施例的结构示意图。 [0040] FIG 5 illustrates an embodiment of the structure of the kernel operating mode memory device receiving data blocks provided by the present invention.

具体实施方式 Detailed ways

[0041] 为使本发明的目的、技术方案和优点更加清楚,下面将结合附图及具体实施例对本发明作进一步的详细描述。 [0041] To make the objectives, technical solutions, and advantages of the invention more clearly, the accompanying drawings and the following specific embodiments of the present invention will be further described in detail.

[0042] 图I为本发明提供的在内核运行态下内存中数据块的发送方法实施例的流程示意图。 [0042] Figure I is a schematic flow diagram of an embodiment in kernel memory operating mode transmission method of the present invention to provide a data block. 图I所示方法实施例包括: As shown in Figure I embodiment includes a method:

[0043] 步骤101、当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间; [0043] Step 101, when detecting the data block of the data packet need unlimited bandwidth link transmission, the receiving side notifies register for storing said data packet in memory all the data blocks;

[0044] 步骤102、为本地用于缓冲所述数据包中全部数据块的内存空间进行注册; [0044] Step 102, a local buffer memory of the data packet to register all the data blocks;

[0045] 步骤103、当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中; All data block [0045] Step 103, after detecting the local and the receiving end to complete the registration memory space, the data packet is filled into a work request;

[0046] 步骤104、通过无限带宽链路传输的向所述接收端发送所述工作请求。 [0046] Step 104, through InfiniBand link transmission to the receiving end sends the work request.

[0047] 本发明提供的方法实施例,将所述数据包的全部数据块通过一个工作请求发送到所述数据包的接收端,与现有技术首先判断是否有页不对齐数据,只传输页对齐数据相比,减少了发送同一数据包中全部数据块的时间,提高了发送的效率;进一步的,也减少了本地进行内存注册的次数,降低了处理器的负担。 [0047] The embodiment of the method provided by the embodiment of the present invention, the entire data block of the data packet transmitted to the receiving end of the packet through a work request, the prior art determines whether the page is first unaligned data transfers only the page compared alignment data, reducing the time to transmit all data blocks in the same data packet, transmission efficiency improves; further, also reduces the number of register memory locally, reducing the burden on the processor.

[0048] 图2为本发明提供的在内核运行态下内存中数据块的接收方法实施例的流程示意图。 [0048] FIG. 2 is a schematic flowchart of an embodiment in kernel mode operating method for receiving data blocks in the memory provided by the present invention. 图2所示方法实施例包括: The method shown in FIG. 2 embodiment includes:

[0049] 步骤201、获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知; [0049] Step 201, get notifications register for storing the data packets in an InfiniBand link transmission memory space comprises all data blocks;

[0050] 步骤202、为本地用于存储所述数据包中全部数据块的内存空间进行注册; [0050] Step 202, the local memory space for storing all the data in the data packet block to be registered;

[0051] 步骤203、在通过无限带宽链路接收到一包括所述数据包中全部数据包的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 After [0051] Step 203, upon receiving a work request comprises a data packet of all packets over an infinite bandwidth link, the data block of the data packet stored in the memory space of the local registration obtained.

[0052] 本发明提供的方法实施例,由于接收端一次性能够接收到数据包中全部数据块,与现有技术中在数据包中有页不对齐的数据块时需要多次接收相比,减少了接收同一数据包中全部数据块的次数,缩短了数据块的接收时间,提高了数据包的传输效率;进一步的,也减少了本地进行内存注册的次数,降低了处理器的负担。 [0052] Example embodiments of the present invention provides, compared to when the need for multiple receiving disposable Since the receiving side can receive all the data packets in the block, the prior art data block is not aligned in the page packet, reducing the number of all the received data blocks in the same data packet, reducing the reception time data block, packet transmission efficiency is improved; furthermore also reduces the number of register memory locally, reducing the burden on the processor.

[0053] 下面以一应用实例对本发明实施例提供的方法实施例作进一步介绍: [0053] In the following example of an application of the present invention will be further described embodiment of the method provided by:

[0054] 图3为本发明提供的在内核运行态下内存中数据块的传输方法应用实例的流程示意图。 [0054] Fig 3 a schematic flow chart in kernel operating mode transmission method is an application example of the memory blocks provided by the present invention. 图3所示方法应用实例包括: Example 3 Application of the method shown in FIG comprising:

[0055] 步骤301、发送端获取需要通过无限带宽链路传输的待发送的数据包。 [0055] Step 301, the sender needs to obtain the data packet to be transmitted unlimited bandwidth link transmission.

[0056] 步骤302、发送端通知该数据包的接收端为自身用于存储该数据包的内存进行注册,包括通知该接收端本次发送的数据包的格式类型。 [0056] Step 302, the receiving end notifies the transmitting end of the data packet to its own memory for storing the data packet registration, including the format type of the packet notifies the receiving end of this transmission.

[0057] 在本步骤中,与现有技术发送具体的格式信息不同的是,本发明发送的是格式标识,其占用的存储空间要小于具体内容所占用的空间。 [0057] In this step, different from the prior art transmission format information is specific, the present invention is the transmission format identification, which storage space is less than space occupied by the specific content. 其中每个格式标识所对应的具体内容可以由发送端和接收端预先协商好。 Wherein each of the specific content corresponding format identification may be previously negotiated by the transmitter and receiver. 在需要使用格式标识时,根据预先设置的格式信息对应的标识,获取所述数据包的格式信息的格式标识;通过向所述接收端发送所述数据包的格式标识通知接收端为用于存储所述数据包中数据块的内存进行注册。 When required format identification, the format information corresponding to the identifier in accordance with a preset acquiring format information of the format identification data packet; receiving end notifies the transmission format that identifies the data packet to the receiving end for storing memory data block in the data packet is registered. 其中所述格式信息为本地和所述接收端预先为每个格式信息的描述信息设置的字符数小于该格式信息的描述信息字符数的标识。 Wherein the format information for the local and the receiving end in advance the number of characters of each description information format information is set to less than the number of format information identifying information of the character described. 由于在通知格式信息时,采用的字段较少的标识代替具体内容,减少了传输的数据量。 Since the notification message format, identifies the less substituted for the specific field of use of the content, the amount of data transferred.

[0058] 当然,该格式类型的信息可转化为直接的格式要求消息,比如所需内存段的个数(即有几个数据块),内存端的大小(即数据块所需的地址空间)以及每个数据块访问时所需的密钥。 [0058] Of course, the format type of information can be directly converted to a format required message, such as the required number of memory segments (i.e., a few blocks of data), the size (i.e., the desired data block address space) and end memory required for accessing each data block key. 特别地,在已知所有数据包最多有两个数据段,其中一个必定是包头的情况下,该类型消息可转化为整个数据包的长度消息或数据包的数据段(非包头段)的长度。 In particular, in all the known data packet has at most two segments, one of which must be the case where the header, the message may be converted to the type (non-header segment) segment length data messages or data packets of the entire packet length .

[0059] 步骤303、发送端在未检测到接收端对所述数据包中数据块的内存注册完成前,为本地用于缓冲所述数据包中数据块的内存进行注册。 [0059] Step 303, the sending end to the receiving end is not detected prior to the registration of memory data in the data packet block is completed, the local memory for buffering data blocks of packet data in the register.

[0060] 其中现有技术中内存注册的方法均适用于本发明,此处不在赘述。 [0060] The method of the prior art wherein the memory register of the present invention are suitable for use herein is not repeated.

[0061] 在本步骤中,触发本地进行内存注册的条件是本地通知接收端进行内存;而现有技术中是将接收到接收端发送的内存注册完成后才触发的,由上可知,本应用实例中发送端注册内存的时间明显早于现有技术的注册时间,从而能够在接收到送的内存注册完成时,已完成本地的内存注册,直接进行数据块发送流程,也进一步缩短了数据块传输所需的时间。 [0061] In this step, the trigger condition register of the local memory for the receiving end notifies the local memory; the prior art and is received into the memory register of the receiving end after completion trigger is transmitted, from the foregoing, this application when the transmitting side register memory instances significantly earlier than the time of registration date prior art, it is possible to send the received memory registration is completed, the completed local memory register, data block transmission process directly, but also to further shorten the data block the time required for transmission.

[0062] 步骤304、接收端为用于存储所述数据包中全部数据块的内存空间进行注册。 [0062] Step 304, the receiving end for storing the data packet in memory all the data block register.

[0063] 步骤305、发送端判断本地和接收端是否都已经完成内存注册操作。 [0063] Step 305, the sending end and the receiving end determines whether the local memory registration operation has been completed.

[0064] 如果已完成,则执行步骤306 ;否则,继续执行步骤305。 [0064] If completed, step 306; otherwise, proceed to step 305.

[0065] 步骤306、发送端将将所述数据包的全部数据块填充到一个工作请求中。 [0065] Step 306, the sender of the data packet to all of the data blocks is filled into a work request.

[0066] 在本步骤中,由于将该数据包的全部数据块通过一个工作请求来发送,那么即使该数据包中右不对齐的首数据块、左不对齐的尾数据块和页不对齐的非首尾数据块中的至少一种,该数据包中的数据块也不会分多次发送。 [0066] In this step, since all data blocks of the data packet transmitted by a work request, even if the packet is not aligned in the right data block header, the left end of the data block is not aligned and unaligned page at least one non-inclusive data block, the data block packet data is not transmitted multiple times.

[0067] 步骤307、发送端通过无限带宽链路向接收端发送所述工作请求。 [0067] Step 307, the sender sends the work request to a receiving end through an infinite bandwidth link.

[0068] 步骤308、接收端将所述数据包的数据块存储到本地注册得到的内存空间。 [0068] Step 308, the data block of the data packet to the receiving side storage space of the local memory register obtained.

[0069] 可选的,在步骤308之后,所述方法应用实例还包括: [0069] Optionally, after step 308, the method further comprises application examples:

[0070] 步骤309、判断是否接收到包括所述接收端成功接收所述数据包中数据块的信息; [0070] Step 309 determines whether the receiving terminal comprises receiving the information of the data packet successfully received in the data block;

[0071] 在本步骤中,通过判断接收端对数据块的接收情况,保证数据块能够成功传输到接收端,保证传输的可靠性。 [0071] In this step, is determined by the receiving side receiving condition of the data block, a data block is possible to ensure successful transmission to the receiving end, to ensure the reliability of transmission.

[0072] 如果接收到,则流程结束;否则,继续执行步骤306。 [0072] If received, the process ends; otherwise, proceed to step 306.

[0073] 本发明提供的方法应用实例,发送端将所述数据包的全部数据块通过一个工作请求发送到所述数据包的接收端,与现有技术首先判断是否有页不对齐数据,只传输页对齐数据相比,减少了发送同一数据包中全部数据块的时间;另一方面,由于接收端一次性能够接收到数据包中全部数据块,与现有技术中在数据包中有页不对齐的数据块时需要多次接收相比,减少了接收同一数据包中全部数据块的次数,缩短了数据块的接收时间,从而达到缩短数据包中全部数据块的传输时间的目的,提高了数据包的传输效率。 [0073] The method of the present invention provides an application example, the receiving end of the data block of all data packets transmitted by a work request packet to the transmitting end, the prior art is first determined whether the page data is not aligned, only compared page-aligned data transmission, reducing the time to transmit all data blocks in the same data packet; on the other hand, since the receiving end can receive a one-time all the data packets in the data block, the prior art there are pages in the data packet require multiple unaligned receives data blocks reduced as compared to the number of all the received data blocks in the same data packet, reducing the reception time data block, so as to achieve the purpose of shortening the transmission time all the packets of data blocks, to improve transmission efficiency of the packet.

[0074] 需要说明的是,本发明提供的方法可以应用在装有IB卡的服务器上,且服务器内部采用Infiniband连接代替总线连接,该服务器中运行的程序通过驱动程序提供的接口基于底层物理传输协议采用系统中的内核线程传输数据块。 [0074] Incidentally, the present invention provides a method may be applied on the server with the card IB, and the server uses Infiniband connection instead the internal bus, the interface program running on the server provided by the driver based on the underlying physical transport protocol uses the kernel thread transport block system. 当然,也不限于此,可以应用在其他设备上基于内核态下传输数据块的场景。 Of course, it is not limited thereto and may be applied to other devices in a scene based on the transmission block of data kernel mode.

[0075] 图4为本发明提供的在内核运行态下内存中数据块的发送装置实施例的结构示意图。 [0075] FIG. 4 configuration diagram of an embodiment in kernel memory apparatus operating mode data transmission blocks provided by the present invention. 结合图I和图3所述的方法,图4所示的装置实施例包括: Method I in conjunction with FIGS. 3 and the device shown in FIG. 4 embodiment includes:

[0076] 通知模块401,用于当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间; When [0076] The notification module 401, a data block is detected when the packet need unlimited bandwidth link transmission, the receiving side notifies register for storing said data packet in memory all the data blocks;

[0077] 注册模块402,与所述通知模块401相连,用于为本地用于缓冲所述数据包中全部数据块的内存空间进行注册; [0077] The registration module 402, the notification module 401 is connected, for the data packet to buffer memory all the data blocks to local register;

[0078] 填充模块403,与所述注册模块402相连,用于当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中; [0078] filling module 403, connected to the registration module 402, for all detected data block when the receiving end and the local memory space to complete registration, the packet is filled into a work request;

[0079] 发送模块404,与所述填充模块403相连,用于通过无限带宽链路向所述接收端发送所述工作请求。 [0079] The transmitting module 404, 403 is connected to the filling module, for transmitting said work request to the receiving end by an infinite bandwidth link.

[0080] 其中所述通知模块401,还用于通知所述接收端所述数据包中全部数据块的格式信息,其中所述格式信息为本地和所述接收端预先为每个格式信息的描述信息设置的字符数小于该格式信息的描述信息字符数的标识。 [0080] wherein the notification module 401 is further configured to format information in the data packet all of the data blocks to be notified to the receiving terminal, wherein the format information describes the local and the pre-format information for each receiving end the identification number is smaller than the number of characters provided description information of the character format information.

[0081] 其中所述注册模块,用于在未检测到接收端对所述数据包中数据块的内存注册完成前,为本地用于缓冲所述数据包中全部数据块的内存空间进行注册。 [0081] wherein the registration module is configured to prior to the reception side does not detect the registration memory of the data packet block is completed, a local buffer memory of the data packet to register all the data blocks.

[0082] 本发明提供的装置实施例,将所述数据包的全部数据块通过一个工作请求发送到所述数据包的接收端,与现有技术首先判断是否有页不对齐数据,只传输页对齐数据相比,减少了发送同一数据包中全部数据块的时间,提高了发送的效率;进一步的,也减少了本地进行内存注册的次数,降低了处理器的负担。 [0082] The present invention provides embodiments of all of the data block of the data packet through the receiving side transmits a work request to the data packet, the prior art determines whether the page is first unaligned data transfers only the page compared alignment data, reducing the time to transmit all data blocks in the same data packet, transmission efficiency improves; further, also reduces the number of register memory locally, reducing the burden on the processor. [0083] 图5为本发明提供的在内核运行态下内存中数据块的接收装置实施例的结构示意图。 [0083] FIG 5 illustrates an embodiment of the structure of the kernel operating mode memory device receiving data blocks provided by the present invention. 结合图2和图3所述的方法,图5所示的装置实施例包括: A method in connection with FIG. 2 and FIG. 3, the device shown in FIG. 5 embodiment includes:

[0084] 获取模块501,用于获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知; [0084] The acquisition module 501, configured to obtain data comprising a register for storing the packet through InfiniBand link transmission in memory all the data blocks to be notified;

[0085] 注册模块502,与所述获取模块501相连,用于为本地用于存储所述数据包中全部数据块的内存空间进行注册; [0085] The registration module 502, the acquisition module 501 is connected to the local data packet for storing memory data block all of the registration;

[0086] 存储模块503,与所述注册模块502相连,用于在通过无限带宽链路接收到一包括所述数据包中全部数据快的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 [0086] a storage module 503, the registration module 502 is connected, for after receiving a data packet comprising the data of all the fast work request by unlimited bandwidth link, the data block is stored into the data packet Sign up to get local memory space. [0087] 本发明提供的装置实施例,由于接收端一次性能够接收到数据包中全部数据块,与现有技术中在数据包中有页不对齐的数据块时需要多次接收相比,减少了接收同一数据包中全部数据块的次数,缩短了数据块的接收时间,提高了数据包的传输效率;进一步的,也减少了本地进行内存注册的次数,降低了处理器的负担。 The invention provides apparatus [0087] according to the present embodiment, since the need for multiple receiver compared to when the receiving end can receive a one-time all the data packets in the data block, and data blocks in the prior art are not aligned in the page packet, reducing the number of all the received data blocks in the same data packet, reducing the reception time data block, packet transmission efficiency is improved; furthermore also reduces the number of register memory locally, reducing the burden on the processor.

[0088] 另外,结合图4和图5所示的装置,本发明还提供一种在内核运行态下内存中数据块的传输系统,包括发送装置和接收装置,其中: [0088] In addition, the coupling means shown in FIGS. 4 and 5, the present invention further provides a transmission system of data blocks in memory in the kernel operating mode, comprising transmitting means and receiving means, wherein:

[0089] 所述发送装置,用于当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间;为本地用于缓冲所述数据包中全部数据块的内存空间进行注册;当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中;通过无限带宽链路向所述接收端发送所述工作请求。 [0089] The transmitting means for detecting when the data block packet requires infinite bandwidth link transmission, the receiving side notifies register for storing the data packet all of the memory space of the data block; with local buffering the data packet to all the memory space of the data block register; Upon detecting the local and the receiving end to complete the registration memory space, filling the entire data block of the data packet to a work request; by Infiniband said work request link transmission to the receiving end.

[0090] 所述接收装置,用于获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知;为本地用于存储所述数据包中全部数据块的内存空间进行注册;在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 [0090] The receiving means for obtaining by comprising registration notification packet for storing unlimited bandwidth link transmission of all the data memory block; local memory for storing the entire data block of the data packet space register; after receiving the data packet comprises a work request by all the data blocks unlimited bandwidth link, storing the data block packet into the memory space obtained locally registered.

[0091] 本发明提供的系统实施例,每次传输时以数据包为传输单元,不采用现有技术中以页对齐的数据块为传输单元,不会出现数据包中有页不对齐的数据块时需要分多次传输的问题,减少了数据块的传输次数,缩短了数据块的接收时间,提高了数据包的传输效率;进一步的,也减少了本地进行内存注册的次数,降低了处理器的负担。 [0091] The present invention provides an embodiment, each transmission of packet data transmission unit, the data block is not used in the prior art is page-aligned transmission unit, a data package from the page data does not appear unaligned when the sub-block requires multiple transmissions the problem, reducing the number of transfer data blocks, the shortening of the reception time data block, packet transmission efficiency is improved; furthermore also reduces the number of locally registered memory, which reduces processing the burden is.

[0092] 本领域普通技术人员可以理解上述实施例的全部或部分步骤可以使用计算机程序流程来实现,所述计算机程序可以存储于一计算机可读存储介质中,所述计算机程序在相应的硬件平台上(如系统、设备、装置、器件等)执行,在执行时,包括方法实施例的步骤之一或其组合。 [0092] Those of ordinary skill in the art understand that all or part of the steps of the above-described embodiments may be implemented using a computer program process, the computer program may be stored in a computer-readable storage medium, the computer program in the appropriate hardware platform (such as system, apparatus, equipment, device) to perform, when executed, the method comprising the steps of the one embodiment or a combination thereof.

[0093] 可选地,上述实施例的全部或部分步骤也可以使用集成电路来实现,这些步骤可以被分别制作成一个个集成电路模块,或者将它们中的多个模块或步骤制作成单个集成电路模块来实现。 [0093] Alternatively, all or part of the steps of the above-described embodiments may be implemented using an integrated circuit, these steps may be made into one integrated circuit modules, or by making them into a single step or a plurality of integrated modules circuit module. 这样,本发明不限制于任何特定的硬件和软件结合。 Thus, the present invention is not limited to any particular hardware and software combination.

[0094] 上述实施例中的各装置/功能模块/功能单元可以采用通用的计算装置来实现,它们可以集中在单个的计算装置上,也可以分布在多个计算装置所组成的网络上。 Each device [0094] Examples of the above-described embodiments / function module / function unit general purpose computing devices may be employed to implement, they can be integrated on a single computing device or distributed in a network composed of multiple computing devices.

[0095] 上述实施例中的各装置/功能模块/功能单元以软件功能模块的形式实现并作为独立的产品销售或使用时,可以存储在一个计算机可读取存储介质中。 When [0095] the above-described embodiments each device / function modules / functional unit is implemented as an independent product sold or used in the form of functional software modules may be stored in a computer-readable storage medium. 上述提到的计算机可读取存储介质可以是只读存储器,磁盘或光盘等。 The above-mentioned computer-readable storage medium may be a read-only memory, magnetic or optical disk.

[0096] 以上所述,仅为本发明的具体实施方式,但本发明的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本发明揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本发明的保护范围之内。 [0096] The above are only specific embodiments of the present invention, but the scope of the present invention is not limited thereto, any skilled in the art in the art within the technical scope of the present invention is disclosed, variations may readily occur or Alternatively, it shall fall within the protection scope of the present invention. 因此,本发明的保护范围应以权利要求所述的保护范围为准。 Accordingly, the scope of the present invention should be the scope of the claims and their equivalents.

Claims (9)

1. 一种在内核运行态下内存中数据块的发送方法,其特征在于,应用在装有IB卡的服务器上,且服务器内部采用Infiniband连接代替总线连接,该服务器中运行的程序通过驱动程序提供的接口基于底层物理传输协议采用系统中的内核线程传输数据块,包括: 当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间; 为本地用于缓冲所述数据包中全部数据块的内存空间进行注册; 当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中; 通过无限带宽链路向所述接收端发送所述工作请求。 1. A method of transmitting kernel operating mode of the data memory block, wherein, on the application server with IB card, and the server uses Infiniband connection instead the internal bus, the program running on the server by the driver interface provides the use of kernel threads transport block system based on the underlying physical transport protocol, comprising: when detecting the data block of the data packet need unlimited bandwidth link transmission, the receiving side notifies register for storing said data packet all memory data block; local buffering for the data packet in memory all the data block to be registered; Upon detecting the local and the receiving end to complete the registration memory space, all of the data packets data blocks filled into a work request; transmitting said work request to the receiving end by an infinite bandwidth link.
2.根据权利要求I所述的方法,其特征在于,在所述通知接收端注册用于存储所述数据包中全部数据块的内存空间时还包括: 通知所述接收端所述数据包中全部数据块的格式标识,其中所述格式标识为本地和所述接收端预先为每个格式信息的描述信息设置的字符数小于该格式信息的描述信息字符数的标识。 2. The method of claim I, wherein, in the receiving end notifies all registered memory space for storing data blocks in the data packet it further includes: notifying the receiving end in the data packet all the format identification data block, wherein the local identifier format and the receiving end is smaller than the predetermined number of identification characters of the description information format information for the number of characters of each description information format information provided.
3.根据权利要求I所述的方法,其特征在于,所述为本地用于缓冲所述数据包中全部数据块的内存空间进行注册,包括: 在未检测到接收端对所述数据包中数据块的内存注册完成前,为本地用于缓冲所述数据包中全部数据块的内存空间进行注册。 3. The method of claim I, wherein the means for buffering said data packet in memory all the data block is registered locally, comprising: a receiving end is not detected for the data packet register memory before the data block is completed, the local memory space for buffering said data packet in all the data block register.
4. 一种在内核运行态下内存中数据块的接收方法,其特征在于,应用在装有IB卡的服务器上,且服务器内部采用Infiniband连接代替总线连接,该服务器中运行的程序通过驱动程序提供的接口基于底层物理传输协议采用系统中的内核线程传输数据块,包括: 获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知; 为本地用于存储所述数据包中全部数据块的内存空间进行注册; 在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 4. A method for receiving operating mode kernel memory block, wherein, on the application server with IB card, and the server uses Infiniband connection instead the internal bus, the program running on the server by the driver interface provides the use of kernel threads transport block system based on the underlying physical transport protocol, comprising: a register for storing comprises obtaining notification packet through InfiniBand link transmission in memory all the data blocks; storing a local said packet memory to register all the data blocks; after receiving the data packet comprises a work request by all the data blocks unlimited bandwidth link, storing the data block of the data packet to a local register to give memory space.
5. 一种在内核运行态下内存中数据块的发送装置,其特征在于,应用在装有IB卡的服务器上,且服务器内部采用Infiniband连接代替总线连接,该服务器中运行的程序通过驱动程序提供的接口基于底层物理传输协议采用系统中的内核线程传输数据块,包括: 通知模块,用于当检测到有数据包的数据块通过无限带宽链路需要传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间; 注册模块,与所述通知模块相连,用于为本地用于缓冲所述数据包中全部数据块的内存空间进行注册; 填充模块,与所述注册模块相连,用于当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中; 发送模块,与所述填充模块相连,用于通过无限带宽链路向所述接收端发送所述工作请求。 A data transmission apparatus memory blocks running in kernel mode, wherein, with the application on the server IB card, and the server uses Infiniband connection instead the internal bus, the program running on the server by the driver interface provides the use of kernel threads transport block system based on the underlying physical transport protocol, comprising: a notification module, configured to, when the data block is detected by the data packet transmission unlimited bandwidth link needs to notify the receiving end for storing registration the packet data block all of the memory space; registration module, connected to the notification module for local buffering of the data packets for memory space to register all the data blocks; filling module, and the registration module is connected, for detecting when the receiving end and the local memory space completed registration, filling the entire data block of the data packet to a work request; transmitting module, connected to the filling module, for transmitting said work request to the receiving end by an infinite bandwidth link.
6.根据权利要求5所述的装置,其特征在于: 所述通知模块,还用于通知所述接收端所述数据包中全部数据块的格式标识,其中所述格式标识为本地和所述接收端预先为每个格式信息的描述信息设置的字符数小于该格式信息的描述信息字符数的标识。 6. The apparatus as claimed in claim 5, wherein: the notification module is further configured to identify the format of said receiving terminal notifies all packet data blocks, wherein said format as the local and the receiving end in advance the number of characters of each description information format identification information is set smaller than the number of characters of the description information format information.
7.根据权利要求5所述的装置,其特征在于: 所述注册模块,用于在未检测到接收端对所述数据包中数据块的内存注册完成前,为本地用于缓冲所述数据包中全部数据块的内存空间进行注册。 7. The apparatus as claimed in claim 5, wherein: said register means for receiving is not detected before the end of the memory registration completion packet data blocks, said local data buffering for all memory blocks to register the package.
8. 一种在内核运行态下内存中数据块的接收装置,其特征在于,应用在装有IB卡的服务器上,且服务器内部采用Infiniband连接代替总线连接,该服务器中运行的程序通过驱动程序提供的接口基于底层物理传输协议采用系统中的内核线程传输数据块,包括: 获取模块,用于获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知;· 注册模块,与所述获取模块相连,用于为本地用于存储所述数据包中全部数据块的内存空间进行注册; 存储模块,与所述注册模块相连,用于在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 A running in kernel mode memory device receiving data blocks, characterized in that, on the application server with IB card, and the server uses Infiniband connection instead the internal bus, the program running on the server by the driver It provides the physical interface based on the underlying transport protocol kernel thread transport block system, comprising: an obtaining module, configured to obtain notification packet unlimited bandwidth link transmission data in the entire memory space comprises a register for storing block by; · registration module, connected to the obtaining module is configured for storing the local data packet in memory all the data block to be registered; memory module, connected to the registration module, for infinite bandwidth link by after receiving the data packet comprises a work request all of the data blocks, the data blocks of the data packet stored in the memory space of the local registration obtained.
9. 一种在内核运行态下内存中数据块的接收装置,其特征在于,应用在装有IB卡的服务器上,且服务器内部采用Infiniband连接代替总线连接,该服务器中运行的程序通过驱动程序提供的接口基于底层物理传输协议采用系统中的内核线程传输数据块,包括发送装置和接收装置,其中: 所述发送装置,用于当检测到有数据包的数据块需要通过无限带宽链路传输时,通知接收端注册用于存储所述数据包中全部数据块的内存空间;为本地用于缓冲所述数据包中全部数据块的内存空间进行注册;当检测到本地和所述接收端都完成内存空间的注册后,将所述数据包的全部数据块填充到一个工作请求中;通过无限带宽链路向所述接收端发送所述工作请求; 所述接收装置,用于获取包括注册用于存储通过无限带宽链路传输的数据包中全部数据块的内存空间的通知; A memory device receiving a data block in the kernel operating mode, wherein, with the application on the server IB card, and the server uses Infiniband connection instead the internal bus, the program running on the server by the driver interface provides the use of kernel threads transport block system based on the underlying physical transport protocol, comprising transmitting means and receiving means, wherein: said transmitting means for detecting when the data block packet need unlimited bandwidth link transmission when notifying the receiving side register for storing said data packet in memory all the data blocks; local buffering for the data packet in memory all the data block to be registered; and when detecting that the local receiving end after completion of the registration memory, the entire data block of the data packet is filled into a work request; transmitting said work request to the receiving end by an infinite bandwidth link; said receiving means, for obtaining a registration comprising by storing the notification packet unlimited bandwidth link transmission in memory all the data blocks; 本地用于存储所述数据包中全部数据块的内存空间进行注册;在通过无限带宽链路接收到一包括所述数据包中全部数据块的工作请求后,将所述数据包的数据块存储到本地注册得到的内存空间。 For locally storing said data packet memory to register all the data blocks; after receiving the data packet comprises a work request by all the data blocks unlimited bandwidth link, the packet data block is stored Sign up to get local memory space.
CN 201010546222 2010-11-15 2010-11-15 Method and device for transmitting data block in memory, method and device for receiving data block in memory and transmission system CN102073593B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 201010546222 CN102073593B (en) 2010-11-15 2010-11-15 Method and device for transmitting data block in memory, method and device for receiving data block in memory and transmission system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 201010546222 CN102073593B (en) 2010-11-15 2010-11-15 Method and device for transmitting data block in memory, method and device for receiving data block in memory and transmission system

Publications (2)

Publication Number Publication Date
CN102073593A CN102073593A (en) 2011-05-25
CN102073593B true CN102073593B (en) 2013-01-23

Family

ID=44032139

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 201010546222 CN102073593B (en) 2010-11-15 2010-11-15 Method and device for transmitting data block in memory, method and device for receiving data block in memory and transmission system

Country Status (1)

Country Link
CN (1) CN102073593B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105786624B (en) * 2016-04-01 2019-06-25 浪潮电子信息产业股份有限公司 A kind of dispatching platform based on redis Yu RDMA technology

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6983303B2 (en) 2002-01-31 2006-01-03 Hewlett-Packard Development Company, Lp. Storage aggregator for enhancing virtualization in data storage networks
US7149817B2 (en) 2001-02-15 2006-12-12 Neteffect, Inc. Infiniband TM work queue to TCP/IP translation
CN101303656A (en) 2008-06-06 2008-11-12 中国科学院计算技术研究所 Java application program remote direct memory access method
CN101459676A (en) 2008-12-31 2009-06-17 中国科学院计算技术研究所 Message transmission frame and method based on high-speed network oriented to file system

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7149817B2 (en) 2001-02-15 2006-12-12 Neteffect, Inc. Infiniband TM work queue to TCP/IP translation
US6983303B2 (en) 2002-01-31 2006-01-03 Hewlett-Packard Development Company, Lp. Storage aggregator for enhancing virtualization in data storage networks
CN101303656A (en) 2008-06-06 2008-11-12 中国科学院计算技术研究所 Java application program remote direct memory access method
CN101459676A (en) 2008-12-31 2009-06-17 中国科学院计算技术研究所 Message transmission frame and method based on high-speed network oriented to file system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
InfiniBandSM Trade Association.InfiniBandTM Architecture Specification Volume 1 Release 1.2.《InfiniBandTM Architecture Specification Volume 1 Release 1.2》.2004,全文.

Also Published As

Publication number Publication date
CN102073593A (en) 2011-05-25

Similar Documents

Publication Publication Date Title
CA2509404C (en) Using direct memory access for performing database operations between two or more machines
US7412488B2 (en) Setting up a delegated TCP connection for hardware-optimized processing
EP2602962A1 (en) A method and apparatus to provide data streaming over a network connection in a wireless MAC processor
JP4160642B2 (en) Network data transfer method
EP1059790A2 (en) Apparatus and method for protocol conversion
US8458280B2 (en) Apparatus and method for packet transmission over a high speed network supporting remote direct memory access operations
CN101641683B (en) Protocols and enhanced wireless usb hub
CN105556930B (en) Extended NVMe controller, storage networking and remote access method namespace
US6321269B1 (en) Optimized performance for transaction-oriented communications using stream-based network protocols
US8103813B2 (en) Method and system for hardware based implementation of USB 1.1 over a high speed link
CN102138297B (en) Graphics multi-media ic and method of its operation
US7817634B2 (en) Network with a constrained usage model supporting remote direct memory access
US8176187B2 (en) Method, system, and program for enabling communication between nodes
JP2010532607A (en) Pci Express packet digest of the change of the device, system and method
CN1561490A (en) Enhanced general input/output architecture and related methods for establishing virtual channels therein
JP5902834B2 (en) Explicit flow control for registering the implicit memory
US8296386B1 (en) Method and system for processing network packets
CN104246733B (en) Universal Serial Bus devices for data redirection
CA2432390A1 (en) Method and apparatus for controlling flow of data between data processing systems via a memory
US7966380B2 (en) Method, system, and program for forwarding messages between nodes
JP2001308926A (en) Method and system for transmission of data between receiver and transmitter
CN1638319A (en) Retransmission system and method for a transport offload engine
US8411677B1 (en) Method and system for processing layered networking protocol packets
CN1238796C (en) Device and method for realizing interface conversion
CN105531684B (en) The port general PCI EXPRESS

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C14 Granted