CN102204267A

CN102204267A - Distributed data storage and access systems

Info

Publication number: CN102204267A
Application number: CN2009801330580A
Authority: CN
Inventors: 戴维·C·卡弗尔; 布兰科·J·格罗瓦克
Original assignee: Jungle Network
Current assignee: Jungle Network; Blackwave Inc
Priority date: 2008-07-10
Filing date: 2009-07-09
Publication date: 2011-09-28
Anticipated expiration: 2029-07-09
Also published as: WO2010006134A3; CN102204267B; EP2307991A2; WO2010006134A2

Abstract

A distributed system for content storage and access includes a storage platform having at least a first storage component, and an access platform having one or multiple access components. Each access component is associated with at least one access service. The access service may be an administrative service for receiving a service request initiated by a first user, the service request being associated with a first item of content, and for identifying a characteristic of the service request, a content management service for determining, based at least in part on the characteristic of the service request, a specification of a data transfer operation to be executed in association with the first storage component, and a directory service for maintaining information associated with the first item of content.

Description

Distributed Storage and connecting system

Technical field

The present invention relates to Distributed Storage and connecting system.

Background technology

Storage (for example, can insert by the Internet) based on distributed network has multiple application.A kind of application is the video access.

In the past few decades, along with family expenses high speed internet business is ripe day by day, Online Video stream more and more is subjected to Internet user's welcome.For example when traditional Video transmission system (as cable television system) no longer satisfy the client growing when being convenient to read the demand of instant transmission, the film consumer may turn to the Online Video storage that this service can be provided at once.But, in fact set up a storage capacity the Internet storage system suitable and be used for providing reliable movie download service not a duck soup to the consumer with reasonable prices with the conventional video rental store.Following example has illustrated some difficult problems that this idea exists behind.

Consider such example, a movie download service system wherein has 20K portion film, every length of a film 2 hours, and encoding rate is 2.5Mbps, is configured to serve 15K session simultaneously.In the past, this system can be regarded as a well-equipped large-scale video rental store.(as a comparison, NetFlix has at present approximately above 70K portion films and television programs.) taking about 2.25GB memory space (being 2hr*2.5Mb/s*60sec*60min/8) according to every film, the required hair memory space of storage 20K portion's film is 45TB.In addition, if used the mirror image reduction, storing the required minimum capacity of whole movie library is 90TB, and this system can make up according to for example every 24 hard disks of 4 station servers with the hard disk of about 96 1TB.

In the superincumbent example system, carrying out the required access bandwidth of 15K parallel session is 37.5Gbps, needs each piece in 96 hard disks that the bandwidth of about 400Mbps can both be provided under the situation that the load of hypothesis on each hard disk equates.But such access rate has exceeded the conventional practical rates of general storage system.Under typical operating load condition, traditional storage system can provide the bandwidth of average each disk 50Mbps.Even adjustedly be configured to provide the extraordinary requirement that also can't satisfy example system up to the multimedia storage server of 150～200Mbps bandwidth.And the difficulty that obtains desirable access bandwidth increases gradually along with the scale of example system.

Some determine that for this application the method for system configuration can be from solving the problem of being discussed in essence according to the real-time resource distribution of DYNAMIC DISTRIBUTION, but for large scale system, this problem especially is difficult to solve, because the combination complexity of this class problem enlarges and increases along with system scale usually.In brief, real-time resource distribution of DYNAMIC DISTRIBUTION and scheduling problem can be feature with np complete problem (NP Complete problem) in essence, this means in a time that is traceable to does not exist can calculate deterministic to separate, in other words, these are separated and have the combination complexity in the space and/or on the time.The method of traditional solution np complete problem normally retrains this problem in some way, thereby makes affined problem can obey deterministic separating.But this affined separating can be lost efficacy because of various reasons.For example, restrained separating of problem can't truly be reflected separating of primal problem.And separate when reaching the boundary condition of running into problem space when this is deterministic, this problem can become combinatorial problem again, causes this deterministic separating to produce catastrophic result sometimes.

Summary of the invention

Usually, the present invention has characterized a kind of distributed system that is used for storage and access on the one hand, this system comprises storage platform and access platform, this storage platform has first storage component at least, this access platform has one or more access parts, each inserts parts and all is associated with in a plurality of access services at least one, these access services comprise: management service, be used to the feature that receives the first Client-initiated service request and discern this service request, this service request is relevant with the content first term; Content management service is used for to the detailed description of small part based on the definite data transfer operation that will carry out that is associated with first storage component of feature of service request; And directory service, be used for safeguarding the information that is associated with the content first term.

Others of the present invention also comprise or several in the following feature.

The detailed description of data transfer operation can comprise the permanent identifier of resolving fully of the element of data transfer operation type identifier and first memory.

Service request can comprise upload request.

Content management service can comprise upload service, and this service is used in response to upload request, distributes the resource of first storage component, thereby determines local resource distribution arrangement, finishes complete resource distribution arrangement according to local resource distribution arrangement then.The detailed description of data transfer operation can comprise the description that complete resource distribution is arranged.

Upload request can comprise first property set of the feature of one or more operands of describing the content first term, and upload service to small part is determined complete resource distribution arrangement based on first group of attribute one or more.

Upload request also can comprise the content first term, and data transfer operation can comprise according to the element of complete resource allocation arrangement with the content first term and storing in the respective element of institute's Resources allocation in the first storage component.

Directory service can safeguard that the element with the content first term stores the description of the result in the respective element of the Resources allocation in the first storage component into.

It is further access operation locating content first term in institute's Resources allocation in first storage component that the result's of the element of memory contents first term description makes it possible to utilize selectivity.

Service request can comprise download request.

Content management service can comprise download service, is used in response to download request, and by mutual with directory service, the existing resource that identification is associated with the content first term is distributed arrangement.Data transfer operation can comprise down operation.

Directory service can be safeguarded the description that existing resource configuration is arranged, and distributes to related between the respective element of first term in the element that comprises the content first term and the first storage component.

Download service can be carried out a group polling, to determine whether to satisfy the condition of initiating down operation.

This group polling can comprise the assessment to following or several: the overall work loading condition of first storage component; Load deviation between the respective element of the first storage component that is associated with the element of first term; The Resource Availability that relates to data transfer operation; And the resource usage level of the various operations that are associated with the content first term in the current distributed system.

When satisfying the initiation condition of data transfer operation, download service can be determined the detailed description of down operation, is included in a group element of selecting in the described element of the described first storage component that will be associated for the described first term that down operation inserts.

First storage component can comprise an above memory device, and this memory device is selected from least one in the following type: disc driver, solid-state memory and nonvolatile memory.

Service request can comprise a plurality of attributes, and this attribute comprises one or more in following: title, data length, data type and one group are described the attribute of feature of one or more operands of a plurality of content items.

State one or more operands and comprise one or more in following: popularity, integrity degree, adaptability, availability, power consumption, geographical concentration degree and repulsion degree and topological concentration degree and repulsion degree.

A plurality of access services also comprise the system call service, are used for coordinating the operation of the one or more execution in a plurality of access services in the distributed system.

This distributed system also can comprise a plurality of access service platforms, and each has one or more access parts.In a plurality of access services that provided by these access parts at least one can be assigned in the example on one or more access service platforms.

Directory service can provide the interface to persistent database, the information that the content that this persistent database is used for storing with distributed system is just being handled is associated.

Management service can provide the service request that receives particular type based on the security means of first subscriber authentication.

Data transfer operation can comprise one or more in the following type: upload, download, delete and move.

Other main aspect comprises other combination of above-mentioned aspect and characteristic, and is expressed as method, mechanism, system, other aspects and characteristic computer program and otherwise expression.

Usually, the feature of one aspect of the present invention is a kind of computer implemented method, comprises received content transmission request, and this request comprises first group of supply attribute, the feature of one or more operands of this attribute description content first term; And contents processing transmission request, store described content first term with the resource in the memory allocated environment.

Various aspects of the present invention can comprise following one or multinomial feature.

The method of contents processing transmission request can comprise first group of supply attribute and plan attribute group are compared, and satisfies one group of plan attribute of one or more operands of content first term with identification; Be the resource in the content first term memory allocated environment, to finish and definite complete resource allocation arrangement based on the local resource allocation arrangement that is associated with one group of plan attribute being discerned; And take measures to make the content first term to be stored in the resource of in storage environment, distributing.

Complete resource distribution arrangement can comprise the description of the destination on the resource in each storage environment that is associated in a plurality of elements with the content first term.

This method also can be included in before the received content transmission request, utilizes knowledge, configuring set of properties and the described local resource allocation arrangement that is associated with each plan attribute group specific to the territory.

The local resource allocation arrangement that this method also can be included in memory utilization set of properties in first database and be associated.

One or more operands of content first term comprise one or more in following: popularity, integrity degree, adaptability, availability, power consumption, geographical concentration degree and repulsion degree and topological concentration degree and repulsion degree.

In the plan attribute group each all has corresponding local resource allocation arrangement.

In the plan attribute group each can be represented a memory utilization, and this memory utilization is that a specific content item based on attributive classification is specified a local resource distribution arrangement.

But the content delivery request is the content first term also, and the method in the resource that makes the content first term be stored in storage environment to be distributed comprises: this content first term is stored in the respective element in the resource of being distributed in storage ring change border.

This method also can comprise to be determined to determine complete resource allocation arrangement based on the local resource allocation arrangement that is associated with one group of plan attribute being discerned.

The method of determining complete resource distribution arrangement can comprise the candidate scheme that generates complete resource allocation arrangement based on local resource allocation arrangement iteration, according to the resource allocation standard quality of this scheme is assessed, and based on the quality assessment result of candidate scheme being selected complete resource distribution arrangement.

The resource distribution standard can distribute with the operating load between resource in the storage environment and be associated.

The resource distribution standard can be associated with potential resource distribution conflict degree degree.

The resource distribution standard can be associated with each the level degree of association assessment in one group of operand of the feature of a plurality of projects of describing this content.

Can be for each the level degree of association assessment in one or more operands at least some of a plurality of projects of described content each.

Select the method for complete resource distribution arrangement to comprise: according to target function of resource distribution standard definition, selected complete resource distribution arrangement makes the target function optimization.

Distribute to distribute the method for the resource in the described storage environment can comprise on the respective resources that the element map of content first term is distributed in the storage environment.

This method also can be included in second database files the element map of the content first term result to the respective resources of being distributed in storage environment.

It is the described content first term in location in the further access operation resource of being distributed in described storage environment that the method for in second database mapping result being filed makes it possible to utilize selectivity.

Resource in the storage environment can comprise one or more memory devices, and these memory devices are selected from least one in the following type: disc driver, solid-state memory and nonvolatile memory.

The content delivery request also can comprise the group of two supply attribute, and this attribute comprises one or more in following: title, data length and data type.

Usually, the feature of one aspect of the present invention is a kind of computer implemented method, and this method comprises that reception inserts the request of the content first term on the resource that is stored in the storage environment, and this request is to be initiated by first user; The existing resource that identification associates the element and the respective element on the resource in the storage environment of content first term is distributed and is arranged; Carry out inquiry, to determine whether to satisfy enabled condition, if satisfy enabled condition, then generate the detailed description of access operation, this access operation is configured to make a group element of selecting in the respective element of this resource in storage environment to be access in.

Various aspects of the present invention can comprise following one to or multinomial feature.

The method of carrying out inquiry can comprise the operating load situation of the resource of estimating memory unit.

The method of carrying out inquiry can comprise estimate with storage environment that the content first term is associated in the respective element of resource between the load deviation.

The method of carrying out inquiry comprises that the availability of the ability of determining first storage component, first storage component are some the set in the respective element of resource of the memory unit that is associated with the content first term.

The ability of first storage component can comprise to small part be the access capability of feature with the maximum data reading rate of first storage component.

First storage component can comprise one or more memory devices, and these memory devices are selected from least one in the following type: disc driver, solid-state memory and nonvolatile memory.

The description that existing resource configuration is arranged can be kept in first database, and this description comprises the expression of a plurality of mappings, each mapping all be stored in storage environment in resource on a plurality of examples of content first term in each be associated.

The method that generates the detailed description of access operation comprises each the quality in assessment and a plurality of mappings that the content first term is associated, and based on the result of quality evaluation, selects one preferably to shine upon in a plurality of mappings according to default standard.

Default standard can be associated with potential resource distribution conflict degree.

The request of access can comprise an attribute of describing the characteristic of content first term.

Generate the attribute that the access operation explanation partly depends on the feature of describing the content first term at least.

This attribute can comprise one group of attribute, and this group of attribute comprises one or more in following: filename, data length, data type and global resource identifier (URI).

This method also can comprise based on the authentication of first user identity passage by safety is received one group of specific access request.

Other main aspect comprises other combination of above-mentioned aspect and feature, and is expressed as method, mechanism, system, other aspects and characteristic computer program and otherwise expression.

According to following description and according to claims, other characteristics of the present invention and advantage are more obvious.

Description of drawings

Fig. 1 is the block diagram of an example distributed data access arrangement.

Fig. 2 is for " uploading " diagram of the flow of event of operating period when storage server is uploaded content when an operated client.

Fig. 3 is the diagram of the flow of event of " download " operating period when inserting client from storage server download content for.

Fig. 4 A is the block diagram of the exemplary upload request of system handles.

Fig. 4 B is the flow chart of Content supply flow process.

Fig. 5 shows the example chart that is associated with a plurality of attributes.

Embodiment

1. system survey

With reference to Fig. 1, distributed system 100 is made up of the computer platform that is used for content stores and transmission (delivery) of three kinds of fundamental types: stores service applicator platform 130, access server platform 120 and customer platform 110.Term " platform " is used in reference to the set of one group of parts, and they can for example install (host) on same computer or be distributed on many computers.

Stores service applicator platform 130 can be installed a plurality of storage servers 132, and wherein each storage server coordinated the visit to one group of storage resources in turn, and the reading and writing data ability is provided.For example each server can be served as by a computer, and perhaps its function self is distributed on a plurality of hardware elements.Each storage server can comprise one or more various types of memory devices.The embodiment of memory device comprises disc driver, solid-state memory (such as RAM) and nonvolatile memory (such as flash memory).

Access server platform 120 provides access service, and this access service provides one or more cut-in methods about the data/content of the storage server of system jointly.Some access services come deal with data according to predefined access protocol/system/interface (for example HTTP, RSTP, NFS etc.).Other access service management of system resource are also correspondingly adjusted content and are inserted.The service of management of system resource comprises distributing system resource for example with storage and the Content supply service 124 that transmits content, and allows the permission of session to serve 126 when various conversation request is asked in by system.

Usually, one side of access server platform 120 is physically limited by distributed network 160, opposite side is physically limited by access network 170, wherein, distributed network 160 provides interconnection between stores service applicator platform 130 and access server platform 120, access network 170 provides interconnection between customer platform 110 and access service platform 120.Note, also can carry out communication by distributed network 160 or by other network in the system by the different access service that one or more access server platforms provide.Access network 170 comprises operational network 140 and transmits network 150, wherein, one or more operated client 112 (for example, the keeper in Online Video shop) carries out communication by operational network 140 and access server platform 120, one or more access clients 114 (for example, the client in Online Video shop) are carried out communication by transmitting network 150 with access server platform 120.

Access server platform 120 can hold (host) one or more management servers that can insert from the outside 121, this management server provides the interface that operational network 140 is linked to the service of access server platform 120 and stores service applicator platform 130, for example, provide safety to obtain, retrieve and verify the method that is stored in intrasystem content.By management server 121, operated client 112 can be with the administrator right connecting system, for example: 1) upload content to system; 2) from the system downloads content; 3) existing content or the like in the deletion system.In certain embodiments, management server 121 utilizes http protocol to provide software interface to operated client, the access of these administrator rights can correspond respectively to HTTP's " PUT ", " GET " and " DELETE " method, or the combination of other HTTP method, for example " POST " and " DELETE ".Preferably, access-in management server 121 is given the privilege of operated client 112.

Access server platform 120 also holds one or more Download Server that can insert from the outside 122.By Download Server 122, inserting client 114 can for example flow out data or download content to system request according to one group that the defines rule (system or agreement) that be used for outflow (stream) and download content.

As shown in Figure 1, between memory and client, need not direct the link, because storage separates (decouple) in system with access function.When operating, client executing (for example uploads or downloads content), on behalf of its client and storage server, management server 121 or Download Server 122 carry out communication and mainly data is put into storage resources or obtained data from storage resources, and communicates with for example locating resource and control authority with other elements in the system.This storage has brought convenience with the configuration and the upgrading that separate to system of access, and the system that greatly reduces can be tending towards making the distributed system collapse along with the expansion of system scale synchronously for synchronous dependence.

After receiving upload request from operated client 112, management server 121 receives and comprises input and the content itself (for example, the binary data of encoded video streams) of uploading content external name (URI for example, that is, global resource identifier).When after insert client 114 or operated client 112 and receive download request, Download Server 122 or management server 121 receive the external name that will download content (for example, or URI) respectively, if possible, then provide the content of being asked.

In certain embodiments, management server 121 can be from operated client with uploading or download request is obtained (for example, with request receive) group management sever 121 employed attribute when service is provided to this request.This attribute can comprise for example supplies attribute, switch-in attribute and management attribute.Be maximum concurrent the fail to be convened for lack of a quorum words of an exemplary embodiment that the attribute that provides is provided for being allowed for certain content segment (for example, title or object).For example, if content constraints is maximum 5 concurrent sessions, then the Content supply service can determine that one is arranged preferably for this segment data on the storage server, if surpassed the restriction of 5 sessions subsequently, management service can be refused download request.

Object directory service 128 provides that () interface for example, object directory 129, this persistent database is used for storing the information relevant with the object of system handles to persistent database.Embodiment that can canned data is a distribution condition information of distributing to the external name of internal object identifier (OID), and second embodiment of this information is the information of describing the memory location of content in the storage server/storage resources of system.Detailed second embodiment that illustrate, object directory 129 can comprise a chart that is associated with certain movie name, this chart is described the data of this film and how to be stored on storage server, the number of copies that comprises this film of preservation, and each copy disc driver and the particular sector that are placed.The 3rd embodiment is that expression (is for example supplied attribute, the peak value of content accumulation access bandwidth and accommodation factor), management attribute (for example, whether management allows to create attribute about the session of certain content) and the information of switch-in attribute (for example, management is with the attribute of much speed rates about the session of certain content).

Should be understood that the data that are stored in the object directory are not limited only to these embodiment, but can be any kind and/or quantity data, comprise the total data of the object of external reference.

In following part, two kinds of main operations---the uses of detailed description supply and management service in the environment of " uploading " and " download " that will be in distributed system 100.

2. upload " operation embodiment

With reference to Fig. 2, in the embodiment of " uploading " operation 200, the request 210 that operated client 112 is uploaded film (for example " Shrek I ") by management server 121 initiations to system.Except that request 210, the data that management server also receives comprise the external name and the movie contents itself of content " Shrek I ".Note, the one-way communication from an interface to another interface (although its implicit agreement perhaps in fact comprises two-way carrying out shake communication) on the arrow presentation logic among the figure, double-lined arrow represents that actual data content is transmitted by this communication.Except external name and movie contents, client can also provide attribute, comprise that traditional attribute (for example, content size and content type) and some influence the non-traditional attribute of each contents processing (such as the concurrent access session number of maximum, popularity, adaptedness, integrity degree, life cycle and topological sum geographical attribute).

After receiving external name, management server sends this external name in order 212, and request object directory service 128 is that " Shrek I " creates an object, and shines upon for this external name generates OID.In case successfully create object, object directory service 128 just transmits message 214 and notifies management server 121.Subsequently, management server is by asking 216 to call Content supply service 124 content is provided.Equally, except that request 216, management server also transmits the supply attribute of content, and decision is available during Content supply gives this content with which part system resource allocation so if this attribute is used for.

In the system configuration process, can pre-determine one group of potential resource configuration for content with some type supply attribute, wherein, each allocation list example is as carrying out the copy of content the method for segmentation between one or more storage resources in system.This makes the Content supply service can make supply decision efficiently.For example, the Content supply service obtains and assesses best mapping by these attributes, and for example by the occupation rate of approximate server at given time, and acquisition can be used to satisfy the possible server group of the content demand (for example, performance requirement) of content.

At this, mapping is often referred to the information of description at the potential arrangement of data of the object on system's storage resources, for example comprises relevant disk zone on the disk drive group that will use and each driver.If allow best mapping, mean that in this embodiment the Content supply service can distribute enough blocks for this content and not threaten the performance of whole system in selected disk, then the Content supply service sends the message 218 of carrying mapping result to management server.Then, based on this mapping result, management server is written in the block that is distributed on the storage server 132 by one group of instruction 220 with content.In some cases, when needing many parts of copies of content in the system, the Content supply service can be determined best mapping separately for each copy.

In some cases, best mapping may be failed because of resource allocation conflict.For example, in other segments of the content that is provided, a segment may have been occupied the some blocks on the disc driver, and this block is included in the mapping result of " Shrek I " just.In this case, can be " Shrek I " selected mapping again, repeat evaluation process then, up to the mapping of finding permission.Subsequently, management server is written to content in the block that is distributed by mapping or mapping group on the disc driver.

After content delivery finishes, storage server will " be finished (done) ", and signal 222 sends to management server, next, the directory service of management server denoted object is filed (by message 224) to the mapping of finishing, so that object directory can be discerned the position of content for client's follow-up access.When receiving from the object directory service after report upgrades " finishing (the done) " message 226 finish, management server notifying operation client (by " uploading success (upload success) " message 228) " uploading " operation completes successfully.

" download " operation embodiment

With reference to Fig. 3, in the embodiment of " download " operation 300, insert the request 310 that client initiates to download target film (for example " Shrek II ") by Download Server 122.Usually, after the external name of receiving " Shrek II " content, whether Download Server 122 utilizes the service in the access server platform to decide for this requests for content and can maybe should be allowed to, if can maybe should be allowed to, how to obtain this content from storage server.Whether particularly, Download Server 122 uses external name by asking 312 to come allocating object directory service 128, exist to know this object, if exist, then locatees corresponding object value in object directory.If object exists, the object directory service is returned this object value by message 314 to Download Server.This object value comprises the OID of object, partly or entirely the mapping (or part mapping) of one group of storage resources of this content is preserved in description, and other attribute relevant with this content.Subsequently, Download Server 122 calls the download request that allows service 126 to determine whether allowing this content, if allow, how to handle this download request.

Usually, allow service 126 to be configured to allow the combination of the session consistent, and refusal is when the session of system's threat existing session integrality when saturated with the supported operating load of system resource.Therefore, one of permission/refusal upload or download request before, management service is at first checked the behaviour in service and the availability of each system resource.In the embodiment an of execution mode and this " download " operation, inspection/the reservation of resource is undertaken by management service, comprise, for example: 1) check the bandwidth of distributing, can not make the content request that became with the new session of guaranteeing to expect for the existing session that relates to " Shrek II " identical content; 2) check available server resource, so which (for example can offer the best service in possible a plurality of copies of determining content, even the bandwidth of a copy is still available on the heaviest server of burden), and keep the storage server relevant with this copy; 3) check on for the storage server of downloading the reservation that will insert and keep available element resources (for example, disk time and bandwidth).In addition, according to enforcement, also have the resource of many other types to be assigned with and to check.For example, can check the ability of access server/platform, allow service can reply one and show that access server should redirect the client to the message of a different access server.

Insert in a plurality of copies of the content on the storage server 132 which when satisfying download request in decision,, allow services selection to have the copy of minimum average server load usually even the load difference that will organize between server is taken into account.For example, copy with harmonic(-)mean and high deviation may have higher average of appropriateness but the copy of low deviation not as good as one.And be preferably in the Distribution Statistics that comprises in the selection to a certain degree, thereby guarantee can not make underloading resource collapse (this situation may take place) when conversation failure recovers.Sometimes, if the neither one copy has enough resources this moment, this session is rejected.

When allowing service 126 decisions to allow this download request and selecting this content copy when service client is served, its can send to Download Server 122 together with the mapping of selected copy with message 318, and Download Server is set up download session by message 320 definite storage server block on storage server requests and described storage server then.Subsequently, the content of " Shrek II " can be sent to Download Server respectively and continue to be sent to the access client from memory by data circuit 322 and 324.When client received the partial data content of film " Shrek II ", " download " operation was finished.After " download " operation stops or finishes, can be released or be labeled before the recovery that is used for follow-up garbage for its resources reserved.

4. based on the embodiment of the Content supply of attribute service

In the embodiment of above-mentioned " uploading " operation, Content supply service 140 utilizes supplies the best mapping of arranging content on storage server that attribute has found the requirement (for example, performance requirement) that can satisfy content.The embodiment of a performance requirement is an access bandwidth, and it allows the maximum concurrent number of sessions of access content relevant with any given time.How set forth the Content supply service in the following examples by using a special supply attribute---the maximum concurrent words that fail to be convened for lack of a quorum decide the best mapping of supporting the access bandwidth demand.

With reference to Fig. 4 A, for example, system 400 is configured to receive the movie library of about 20K portion film, and each copy of every film takies the memory space of about 500MB, has the client access rate of 2.5Mbps.Supposing the system 400 has the disc driver of 96 1TB, is evenly distributed on four storage servers 432, thereby the memory capacity of 96TB is provided.For each disc driver, for example have the access bandwidth of 100Mbps, then the access capability of system (that is maximum access bandwidth) is 9.6Gbps.(notice that other embodiment may have different movie features and system's storage and access capability.For example, encode, grow 100 minutes film approximately with 2.5Mbps for one one, each copy can take the memory space of about 2GB.Together with mirror image, 20K portion film will take the memory space (for example, 80 1TB disc drivers) of about 80TB.If each disk has the access bandwidth of 400Mbps, then total access bandwidth of system can reach about 32Gbps.)

For every in storehouse film being carried out good resource distribution arrangement, the method that system usually adopts is that distribution of content that the request rate is high is in more disk, so that the access bandwidth summation of this content is enough to support the concurrent words that fail to be convened for lack of a quorum all under the peak utilization.

For example, when management server 421 receives one when memory is uploaded the request 402 of film " Shrek I " from operated client, it also is apprised of for " Shrek I ", in any given time maximum 140 concurrent words that fail to be convened for lack of a quorum can be arranged, illustrate that the peak velocity rate is 350Mbps (supposes film encode with 2.5Mbps) here.Because each disc driver is supported the access rate of maximum 100Mbps, if film is not distributed on minimum 4 disc drivers, the bandwidth of 350Mbps can't reach so.Therefore, the method that the Content supply service distributes to content is with on data map to 4 disc driver, though perhaps this is not best selection, because when this film reaches peak utilization, other data on these 4 drivers will become and can not obtain (example of competing between film).Better choice is that competition between film is taken into account, thereby distributes such data more extensive, such as on 6,12 even 24 disc drivers.Usually, system has one group of such mapping, and each all can satisfy the 350Mbps bandwidth requirement of service content.

One group of mapping with common objective (for example, bandwidth capacity) here is called as planning (schema).For this movie library example, every film has the corresponding planning of (such as, access bandwidth) of a content demand that satisfies this film.

It is very important to calculate the planning that every film can meet the demands, and the difficulty level of calculating increases gradually along with system scale usually.Therefore, in certain embodiments,, can calculate one group of planning in advance,, and be stored in the database, Content supply service 424 can be used for example based on property distribution a kind of prediction or modeling in order to alleviate the burden of Content supply service.This group is calculated good planning in advance and is used for reducing the computing expense, and improves overall system efficiency.

Shown in Fig. 4 B, by using the database that calculates planning in advance, this exemplary systems 400 provides content with following method.At first, when receiving upload request 480, an acceptable planning of satisfying the content demand is sought in Content supply service 424 from planning database in step 482.Next, with mapping that selected planning is associated in, a best mapping is determined in the Content supply service in step 484, for example, have the mapping of lowest service device occupation rate.In following step 486, if can find enough resources on the disc driver of appointment in the best mapping, then Content supply service 424 will be that content is distributed one group of block (488) on each disc driver before content delivery procedure (step 490) beginning.If but Content supply service 424 can't be located enough resources in the best mapping in step 486, its can select another mapping in planning, and repeating step 486 distributes up to a chosen block that carries out of acceptable mapping.

The universal based on the Content supply service of attribute of this use planning has further instruction in the embodiment of uploading shown in Fig. 4 A " Shrek I ".As previously discussed, need the access bandwidth of 350Mbps in this example.Therefore, in this planning database 470, Content supply service 424 will have the planning S3 (shown in table 472) of 400Mbps bandwidth capacity as acceptable planning.This planning S3 illustrates that also for each content that meets this classification, two copies of this content will be stored in the system, and each copy is distributed on 24 disks, and writes on the R3 district of each disk.

Behind the selected S3, can produce a mapping table 474, describe one group and be planned for may shining upon of disc driver.For example, in mapping table 474, M21, M22, M23 and M24 are corresponding to four the quadrature mappings of S3 to memory, and one group of specific disk (such as use disk 1 to 24 in M21) is used in each mapping, and the disk region of appointment (R3).Content supply service each copy for content in the institute relevant with planning S3 might shine upon is sought best mapping, for example passes through to be similar to the occupation rate of server in preset time.In this embodiment, Content supply services selection mapping M21 and M23 store two copies of " Shrek I " respectively.Can visit if find these two mappings, then block distributor 476 is mapped in according to these and finishes the block distribution on disc driver.If the block that distributes can with and be preserved for taking, management service 421 writes content-data by distributed network 460 to storage server 432.

5. the embodiment of a plurality of attributes of using of Content supply service

In some applications, the Content supply service uses a plurality of attributes to seek the best mapping of arrangement content on storage server.The embodiment of these attributes comprises the attribute of describing popularity, adaptability and integrality.

Popularity refers to the expression for the anticipated demand amount of a content.For example, if content is welcome, then may be than higher for the demand of this content, so supply system will attempt to provide this content with enough resources, to satisfy anticipated demand.

Adaptability refers under the situation of component failure the expection service degree of keeping for information object.For example, when obtaining the good object of adaptability, a client can obtain service by another copy that continues on the reading disk B when reading disk A fails.Usually the required adaptability of object is strong more, and contents providing system need be the resource many more (possible words resource kind is also many more) that this object distributes.

Integrality refers to recover the ability of content behind component failure, even the catastrophic inefficacy of system.For example, if a disk complete failure and can't repairing, the content part that is included on this disk can recover from other copies by this content between internal system or the system, also can recover with coding reconstruct method.

According to specific implementation, the Content supply service can be determined the best mapping of memory contents segment, and this mapping can provide satisfied popularity, adaptability and/or integrality for this content segments.

With reference to Fig. 5, some have been shown in planning table 500 planning of in various degree popularity, adaptability and integrality is provided.For example, with reference to planning 4 (S4), in the process of the film of uploading an adaptive to a certain degree low relatively popularity of needs, provide a primary copy to satisfy the demand of popularity.Simultaneously, the pair copy of two these contents of storage is to guarantee still can to obtain this film by two secondary copies under the situation in not break in service user under the situation that reads the primary copy failure.Planning 5 (S5) provide that another kind of to upload low relatively popularity still inessential but need the scheme of the film of high integrality.At this moment, for example high integrality can realize (can increase amount of calculation although it is so) by using senior redundancy encoding.

In some existing systems, adaptability and integrality are normally implicit is bundled together in (for example, RAID 5) in the memory redundancy scheme.In this description, the separation of these two kinds of attributes, and the advantage that the proposition integrality that each content segments is shown and adaptability require is that the permission system can think that every film provides a demand combinations, and uses the combination that satisfies the demands of optimal method.

6. other embodiment of this distributed system

Referring again to Fig. 1, in certain embodiments, the access server platform 120 in the distributed system 100 can also provide the access service 127 of other type except above-described service is provided.Embodiment is interior unit weight supply service, this service can such as occur using unusual or some situation that content is expired under, determine supplying content (comprising the uplink and downlink supply) again.Another example is dispatch service (Management Service), configuration, monitoring, control and the data acquisition of this service in can coherent system.Dispatch service can be followed the tracks of normal and anomalous event, and so that the diagnostic system problem is with unusual when needed, subsequent operation is carried out in other service in the triggering system (for example heavily supply service).

Each service that system provides can be accomplished in several ways.At some is among the embodiment, and service can be tied to (for example server) on the specific platform.Though these services do not possess adaptability for component failure separately, can heavily distribute by service replication and session level and realize system suitability.In certain embodiments, service can be distributed in a plurality of examples of a group platform.This distribution can be by being divided into a kind of degree that can reach service performance and adaptability demand with problem space (for example, by the process flow waterline, perhaps symmetrical concurrent etc. by object).Other is among the embodiment at some, and service can be from a platform transplantation to another.If particularly service is not to realize by distributed mode, if service was lost efficacy on platform or degenerated, transplanting can make serves on other platform that is relocated in the system.

In some implementations, the stores service applicator platform can be being equipped with storage control and for example being configured to provide 10GbE tie-in line speed to the ready-made computer system of the network control card of one group of storage resources (for example, disk or memory) of configuring.Control card both can be developed voluntarily, also can obtain finished product from the third party.In certain embodiments, storage control provides the not limited high access bandwidth for affiliated memory device.The embodiment of memory device comprises disc driver, solid-state memory (such as RAM) and nonvolatile memory (such as flash memory).

In some implementations, access server platform can be a ready-made computer system that is equipped with the network control card that configures, and this network control card is configured to support the access mechanism of memory contents in a series of connecting systems.Access mechanism is provided by access server, and these access servers can adopt various protocols, and its performance and efficient also can have very big-difference.An embodiment of access server is Microsoft Windows Media Server (WMS), and it is widely used as streaming media service on the internet.Another embodiment is a http server, and it uses http protocol that the user is provided the access for content in the system.The 3rd the Flash Media Server that embodiment is Adobed.

Notice that management server and Download Server are two classes in the potential multiple access server.Access server is to provide the outside general class server that inserts according to the object of predefined rule in system.

Though access server (and access service) and storage server (and stores service) have been described to deposit in respectively on access and the storage platform, in some embodiments, the combination of a kind of access and storage server and service or all access and storage server and service can be configured to and be stored on the common platform.

In some embodiments, operational network 140 is isolated with transport net 150.Management server 121 and Download Server 122 can mutual exclusion be operated client 112 respectively and insert client 114 and insert.Download Server 122 can provide characteristic or the statement that is not suitable for operated client, and vice versa.

The purpose that should be understood that foregoing description is explanation, and and do not lie in and limit the scope of the invention, scope of the present invention is limited by the scope of appended claims.Other execution mode is within following claim scope.

Claims

1. distributed system that is used for content stores and access comprises:

Storage platform has first storage component at least; And

Access platform has one or more access parts, and each inserts parts and is associated with in a plurality of access services at least one, and described access service comprises:

Management service is used to receive the first Client-initiated service request, and the feature that is used to discern described service request, and described service request is associated with the content first term;

Content management service is used for to the detailed description of small part based on the definite data transfer operation that will carry out that is associated with described first storage component of described feature of described service request; And

Directory service is used to safeguard the information that is associated with described content first term.

2. distributed system according to claim 1, wherein, the detailed description of described data transfer operation comprises the data transfer operation type identifier, and the permanent complete analysable identifier of the element of described first storage component.

3. distributed system according to claim 1, wherein, described service request comprises upload request.

4. distributed system according to claim 3, wherein, described content management service comprises:

Upload service is used in response to upload request, and the resource of distributing described first storage component is finished complete resource allocation arrangement according to described local resource allocation arrangement then to determine local resource allocation arrangement;

Wherein, the detailed description of described data transfer operation comprises the description that described complete resource allocation is arranged.

5. distributed system according to claim 4, wherein, described upload request comprises first group of attribute of the feature of one or more operands of describing the content first term, and described upload service to small part is determined described complete resource allocation arrangement based on described first group of attribute one or more.

6. distributed system according to claim 5, wherein, described upload request also comprises described content first term, and described data transfer operation comprises according to the element of described complete resource allocation arrangement with described content first term and storing in the respective element of institute's Resources allocation in the described first storage component.

7. distributed system according to claim 6, wherein, described directory service safeguards that the element with described content first term stores the description of the result in the respective element of the Resources allocation in the described first storage component into.

8. distributed system according to claim 7, wherein, it is that described content first term is located in further access operation in institute's Resources allocation in described first storage component that the result's of the element of memory contents first term description makes it possible to utilize selectivity.

9. distributed system according to claim 1, wherein, described service request comprises download request.

10. distributed system according to claim 9, wherein, described content management service comprises:

Download service is used in response to described download request, and by mutual with described directory service, the existing resource that identification is associated with described content first term is distributed arrangement;

Wherein, described data transfer operation comprises down operation.

11. distributed system according to claim 10, wherein, described directory service safeguards that described existing resource distributes the description of arranging, and distributes to related between the respective element of described first term in the element that comprises described content first term and the described first storage component.

12. distributed system according to claim 11, wherein, a group polling is carried out in described download service, to determine whether to satisfy the condition of initiating described down operation.

13. distributed system according to claim 12, wherein, a described group polling comprises one in following or several assessment: the overall work loading condition of described first storage component; Load deviation between the described respective element of the described first storage component that is associated with the described element of described first term; The Resource Availability that relates to described data transfer operation; And the resource usage level of the various operations that are associated with the content first term in the current described distributed system.

14. distributed system according to claim 13, wherein, when satisfying the condition of initiating described data transfer operation, the described detailed description of described down operation is determined in described download service, is included in a group element of selecting in the described element of the described first storage component that will be associated for the described first term that down operation inserts.

15. distributed system according to claim 1, wherein, described first storage component comprises one or more memory devices, and described memory device is selected from least one in the following type: disc driver, solid-state memory and nonvolatile memory.

16. distributed system according to claim 1, wherein, described service request comprises a plurality of attributes, and described attribute comprises one or more in following: title, data length, data type and one group are described the attribute of feature of one or more operands of a plurality of content items.

17. distributed system according to claim 16, wherein, described one or more operand comprises one or more in following: popularity, integrity degree, adaptability, availability, power consumption, geographical concentration degree and repulsion degree and topological concentration degree and repulsion degree.

18. distributed system according to claim 1, wherein, described a plurality of access services also comprise the system call service, are used for coordinating the one or more performed operation in described a plurality of access services in the described distributed system.

19. distributed system according to claim 1, also comprise a plurality of access service platforms, each has one or more access parts, wherein, at least one in the described a plurality of access services that provided by described access parts is assigned in a plurality of examples on one or more access service platforms.

20. distributed system according to claim 1, wherein, described directory service provides the interface to persistent database, and described persistent database is used to store the information that the content just handled with described distributed system is associated.

21. distributed system according to claim 1, wherein, described management service provides the service request that receives particular type based on the security means of first subscriber authentication.

22. distributed system according to claim 1, wherein, described data transfer operation comprises one or more in the following type: upload, download, delete and move.

23. a computer implemented method comprises:

Received content transmission request, described request comprise first group of supply attribute, the feature of one or more operands of described attribute description content first term; And

Handle described content delivery request, store described content first term with the resource in the memory allocated environment.

24. computer implemented method according to claim 23 wherein, is handled described content delivery request and is comprised:

First group of supply attribute compared with the plan attribute group, satisfy one group of plan attribute of one or more operands of described content first term with identification;

For described content first term is distributed resource in the described storage environment, to finish and definite complete resource allocation arrangement based on the local resource allocation arrangement that is associated with described one group of plan attribute of being discerned; And

Take measures to make described content first term to be stored in the resource of in described storage environment, being distributed.

25. computer implemented method according to claim 24, wherein, described complete resource allocation arrangement comprises the description of the destination on the described resource in each described storage environment that is associated in a plurality of elements with described content first term.

26. computer implemented method according to claim 24 also comprises:

Before receiving described content delivery request, utilize knowledge, the described local resource allocation arrangement of disposing described plan attribute group and being associated with each described plan attribute group specific to the territory.

27. computer implemented method according to claim 26 also comprises:

The local resource allocation arrangement of in first database, storing described plan attribute group and being associated.

28. method according to claim described 23, wherein, one or more operands of described content first term comprise one or more in following: popularity, integrity degree, adaptability, availability, power consumption, geographical concentration degree and repulsion degree and topological concentration degree and repulsion degree.

29. computer implemented method according to claim 24, wherein, each in the described plan attribute group all has corresponding local resource allocation arrangement.

30. computer implemented method according to claim 24, wherein, the memory utilization of each expression in the described plan attribute group, described memory utilization is that the specific content item based on attributive classification is specified local resource allocation arrangement.

31. computer implemented method according to claim 24, wherein, described content delivery request also comprises described content first term, and wherein, the method for taking measures described content first term is stored in the resource of being distributed in described storage environment comprises:

The element of described content first term is stored in the respective element in the resource of being distributed in described storage environment.

32. computer implemented method according to claim 24 also comprises:

Based on the described local resource allocation arrangement that is associated with described one group of plan attribute of being discerned, determine described complete resource allocation arrangement.

33. computer implemented method according to claim 32 wherein, determines that described complete resource allocation arrangement comprises:

Arrange iteration to generate the candidate scheme that described complete resource allocation is arranged based on described local resource allocation;

Quality according to the described scheme of resource allocation criterion evaluation; And

Based on the quality assessment result of described candidate scheme, select described complete resource allocation arrangement.

34. computer implemented method according to claim 33, wherein, the operating load in described resource allocation standard and the described storage environment between resource distributes and is associated.

35. computer implemented method according to claim 33, wherein, described resource allocation standard is associated with potential resource distribution conflict degree.

36. computer implemented method according to claim 33, wherein, the level degree of association assessment of each in one group of operand of described resource allocation standard and the feature of a plurality of projects of describing described content is associated.

37. computer implemented method according to claim 36, wherein, for each the level degree of association assessment in one or more operands be at least some of a plurality of projects of described content each.

38. computer implemented method according to claim 33 wherein, selects described complete resource allocation arrangement to comprise:

According to target function of described resource allocation standard definition,

Wherein, selected complete resource allocation arrangement makes described target function optimization.

39. computer implemented method according to claim 24 wherein, distributes the resource in the described storage environment to comprise:

With the element map of described content first term to the respective resources of in described storage environment, being distributed.

40., also comprise according to the described computer implemented method of claim 39:

In second database, the element map of the described content first term result to the respective resources of being distributed in described storage environment is filed.

41. according to the described computer implemented method of claim 40, wherein, in second database, the described result of mapping filed that to make it possible to utilize selectivity be the described content first term in location in the other access operation resource of being distributed in described storage environment.

42. computer implemented method according to claim 23, wherein, resource in the described storage environment comprises one or more memory devices, and described memory device is selected from least one in the following type: disc driver, solid-state memory and nonvolatile memory.

43. computer implemented method according to claim 23, wherein, described content delivery request also comprises second group of supply attribute, and described attribute comprises one or more in following: title, data length and data type.

44. a computer implemented method comprises:

Receive the request that inserts the content first term on the resource that is stored in the storage environment, described access request is initiated by first user;

The existing resource that identification associates the element and the respective element on the described resource in the described storage environment of described content first term is distributed and is arranged;

Carry out inquiry to determine whether to satisfy enabled condition; And

If satisfy enabled condition, then generate the detailed description of access operation, this access operation is configured to make a group element of selecting in the respective element of the described resource in described storage environment to be access in.

45. according to the described computer implemented method of claim 44, wherein, the described inquiry of described execution comprises:

Estimate the operating load situation of the described resource of memory unit.

46., wherein, carry out described inquiry and comprise according to the described computer implemented method of claim 44:

Estimate with described storage environment that described content first term is associated in the respective element of described resource between the load deviation.

47. according to the described computer implemented method of claim 44, wherein, described execution inquiry comprises:

Determine the availability of the ability of first storage component, described first storage component is some the set in the respective element of described resource of the described memory unit that is associated with described content first term.

48. according to the described computer implemented method of claim 47, wherein, the described ability of first storage component comprise to small part be the access capability of feature with the maximum data reading rate of described first storage component.

49. according to the described computer implemented method of claim 47, wherein, described first storage component comprises one or more memory devices, and described memory device is selected from least one in the following type: disc driver, solid-state memory and nonvolatile memory.

50. according to the described computer implemented method of claim 44, wherein, described existing resource distributes the description of arranging to be stored in first database, described description comprises the expression of a plurality of mappings, each mapping all be stored in described storage environment in described resource on a plurality of examples of described content first term in each be associated.

51. according to the described computer implemented method of claim 50, wherein, the described detailed description that generates described access operation comprises:

The quality of each in assessment and a plurality of mappings that described content first term is associated;

Based on the result of quality evaluation, in described a plurality of mappings, select a preferred mapping according to default standard.

52. according to the described computer implemented method of claim 51, wherein, described default standard is associated with potential resource distribution conflict degree.

53. according to the described computer implemented method of claim 44, wherein, described access request comprises the attribute of the feature of describing described content first term.

54. according to the described computer implemented method of claim 53, wherein, the detailed description of described generation access operation partly depends on the attribute of the feature of describing described content first term at least.

55. according to the described computer implemented method of claim 53, wherein, described attribute comprises one group of attribute, described one group of attribute comprises one or more in following: filename, data length, data type and global resource identifier (URI).

56., also comprise according to the described computer implemented method of claim 44:

Based on the authentication of described first user identity passage by safety is received one group of specific access request.