US20030204602A1 - Mediated multi-source peer content delivery network architecture - Google Patents

Abstract

A multi-source peer content distribution system transfers content files from multiple, distributed peer computers to any requesting computer. The content distribution network coordinates file transfers through a mediation system including a content catalog and a host broker system. The content catalog contains an identification of each content file, the segmented subunits of each file, and the peer caches to which the subunits have been distributed. The host broker system receives content file requests issued over a network from requesting computers. In response, manifest files identifying the request corresponding content subunits and distributed cache locations are returned. The requesting computers can then retrieve and assemble the corresponding content subunits from the peer computers to obtain the requested content file.

Description

BACKGROUND OF THE INVENTION
• 1. Field of the Invention[0001]
• The present invention is generally related to network-based content delivery systems and, in particular, to a streaming media content delivery system supporting multiple, concurrent, peer-based sources of multimedia content accessible subject to central mediation[0002]
• 2. Description of the Related Art[0003]
• The desire for high-quality, on-demand delivery of streaming multimedia and other rich digital content is a principal driving force in the continued development of the broadband Internet infrastructure. Indeed, with the growth of broadband connections, the number, scale, and diversity of multimedia content servers has rapidly increased. Streaming audio and video files, including entertainment, news broadcasts, and instructional programming are now sourced by a variety of mainstream Internet sites. Content delivery through streaming media is broadly recognized as one of the fastest growing technologies related to the Internet.[0004]
• Despite the growth in interest and use, conventional content streaming systems have not been cost-effective or particularly reliable in delivering high-quality content. Streaming media content, including in particular high-quality audio and video, is naturally bandwidth intensive and fundamentally sensitive to varying delivery latencies. Whether due to transient transport overloads, functional interruptions in the network infrastructure, or bandwidth limitations of a content source site, the result is uniformly perceived by a recipient as a reduction in the quality of service of the content source site.[0005]
• Because of the open and shared nature of the Internet, few practical mechanisms can ensure the uninterrupted delivery of broadband content, typically consisting of multi-megabyte files, over the entire delivery path from a source site to a recipient. Known schemes include the use of network edge caches distributed at strategic locations within the network infrastructure controlled by an individual service provider. These network edge caches can be operated to significantly reduce the network traffic through the local network space of the individual service provider. Large edge caches are naturally required to store any significant amount of streaming media content. Implementing a useful number of adequately scaled, geographically distributed edge caches requires a large capital infrastructure investment.[0006]
• Quality of service issues within the domain of individual content source sites are relatively easier to manage. Over the past few years, highly scaled, geographically distributed and even multiply redundant content source system architectures have been developed. Conventionally, these very large-scale systems are considered a baseline requirement to ensure a consistent high quality of service from the source sites. These sites typically employ large-scale server farms, hosting extensive libraries of archived multimedia content, that cumulatively provide sufficient throughput to enable real-time responsiveness and continuous on-demand delivery. Very high-bandwidth Internet connections with sufficient capacity to accommodate peak-demand content access requirements are also required.[0007]
• Unfortunately, conventional content delivery networks, including fully scaled content server systems and extensive edge cache networks, have not proven adequate to broadly ensure a high quality of service to all potential users of the systems. Ultimately, any media content must be delivered as an effectively continuous stream of multimedia data to the recipient computer. The continuity of the stream must remain within the buffer length tolerance supported by the media player on the recipient computer. Transient bandwidth bottlenecks can certainly occur anywhere beyond the scope of a conventional content delivery network. Bottlenecks and delivery latencies can occur even within the network, particularly whenever the stream data is not immediately available in a locally accessible edge cache. Such bottlenecks in the Internet infrastructure are unfortunately both common and unpredictable.[0008]
• Transient bandwidth bottlenecks can also occur in within the content server system itself. The rate of content access requests is highly variable with unpredictable demand peaks. Whenever the access rate exceeds the capabilities of the content server system, connection requests, including ongoing streaming data transfers, are dropped or delayed. Whether due to network or server bottlenecks, the resulting latencies and gaps in the delivery of stream data packets ultimately to the recipient are uniformly seen as source-site quality of service failures.[0009]
• Expanding the conventional content distribution networks to prevent significant transient bandwidth bottlenecks is generally recognized as not practical. Due to the size and diversity of the Internet and the growing demands for streaming content delivery, significantly expanding the edge cache network coverage and the capacity of all included edge caches and streaming media source sites is simply not cost-effective. Furthermore, the costs associated with high-bandwidth Internet access and server throughput grow proportional to peak access demands, which is disproportionately greater than the growth of average access demands. Conventionally, a minimum of 50 percent additional access and server bandwidth is required, if not more, to meet peak bandwidth requirements. This additional bandwidth, however, is unused typically in excess of 90 percent of the time. The capital and operating cost of this additional bandwidth is therefore not directly recoverable. Consequently, content sites and the content delivery network operators have been severely limited in being able to consistently and profitably deliver streaming media content with a high quality of service.[0010]
• Consequently, there is a clear need for a content delivery network architecture that can reliably provide a high quality of service to content stream recipients and that is cost-effective to operate.[0011]
SUMMARY OF THE INVENTION
• Thus, a general purpose of the present invention is to provide an efficient peer-to-peer content distribution network system architecture capable of efficiently providing a high quality of service in the delivery of multimedia data streams to end-users.[0012]
• This is achieved in the present invention through a multi-source peer content distribution system transfers content files from multiple, distributed peer computers to any requesting computer. The content distribution network coordinates file transfers through a mediation system including a content catalog and a host broker system. The content catalog contains, directly or indirectly, an identification of each content file, the segmented subunits of each file, and the peer caches to which the subunits are distributed. The host broker system receives content file requests issued over a network from requesting computers. In response, manifest files identifying the request corresponding content subunits and distributed cache locations are located and returned to the requesting computer. The requesting computers can then retrieve and assemble the corresponding content subunits from the peer computers to obtain the requested content file.[0013]
• An advantage of the present invention is that content is redundantly distributed in the form of discrete segments throughout a peer storage network, permitting retrieval of segments on a best quality-of-service basis determined relative to each computer system that requests a streaming media content file. Multi-source segmented delivery of content also distributes the transport load over multiple content sources while ensuring the availability of multiple sources for all segments. The perceived quality-of-service is both increased and reliably maintained.[0014]
• Another advantage of the present invention is that centralized mediation of segmented file transfers permits strategic planning of the ongoing segmented file transfer load distribution. Central mediation combined with distributed segmented file storage enables the aggregate bandwidth of the content distribution network to be optimally utilized. The complexity and cost of the central content mediation system, including the scale of the network access connections to accommodate worst case usage requirements, are greatly reduced.[0015]
• A further advantage of the present invention is that the mediation system can perform predictive seeding of the content delivery network and adaptive modification of segment distribution in response to changing content file demands. Historical demand patterns, peer node availability and bandwidth capabilities can be used to guide the strategic distribution of content segments throughout the content delivery network. Planned, periodic updates of the content distribution network segment caches can be used to pre-deliver content segments to multiple strategically selected network caches during off-hours, thus minimizing both seeding and subsequent end-user demand spikes.[0016]
• Still another advantage of the present invention is that the content distribution network can include a multi-tiered hierarchy of content segment caches, including peer cache nodes, primary content distribution nodes, and seeding servers. Reliable access to content segments is ensured by wide distribution of the content segments within the segment cache storage tiers and across multiple tiers.[0017]
• Yet another advantage of the present invention is that the central mediation system can be persistently connected to and monitor the state of the available, active peer nodes of the content distribution network. Changes in the availability and supported bandwidth of nodes within the content distribution network are dynamically detected and factored into the ongoing tactical utilization of the content distribution network as mediated by the central server system.[0018]
• Still another advantage of the present invention is that the distribution of content segments is actively maintained by the mediation server system. The distribution of content segments within the content delivery network, which is continuously subject to redistribution as a consequence of content use requests, is tracked and managed by the mediation server system to strategically adapt the distribution pattern to optimally match demand patterns.[0019]
• A yet further advantage of the present invention is that proprietary content is continuously protected by a combination of encryption and digital signatures applied to the content files and to the individual content file segments. The mediation server system maintains the integrity of the content file segments throughout the operations of file segment transport, cache storage, and streaming file assembly and playback. The integrity of content within the content distribution system is thus ensured by the management function of the mediation server system.[0020]
BRIEF DESCRIPTION OF THE DRAWINGS
• These and other advantages and features of the present invention will become better understood upon consideration of the following detailed description of the invention when considered in connection with the accompanying drawings, in which like reference numerals designate like parts throughout the figures thereof, and wherein:[0021]
• FIG. 1 is a block diagram of a content distribution network organized in accordance with a preferred embodiment of the present invention;[0022]
• FIGS. 2A and 2B provide detail views of the segmentation of a streaming media content file in accordance with preferred embodiments of the present invention;[0023]
• FIG. 3 is a block diagram of a peer network client system constructed in accordance with a preferred embodiment of the present invention;[0024]
• FIG. 4 is a block diagram of a peer network mediation server system constructed in accordance with a preferred embodiment of the present invention;[0025]
• FIG. 5 is a flow diagram of a streaming media content file transfer executed with respect to a peer network client system in accordance with a preferred embodiment of the present invention; and[0026]
• FIG. 6 provides a detailed flow diagram of the adaptive request process implemented in a peer network client system constructed in accordance with a preferred embodiment of the present invention.[0027]
DETAILED DESCRIPTION OF THE INVENTION
• The content distribution network (CDN) of the present invention provides a comprehensive system solution to delivering streaming media and other digital content files to end-user systems with a consistent, high quality of service. The end-user systems participate in a distributed network of peer computer systems, organized into a tiered set of content sources, that store and, on request, selectively forward content to any other peer computer system within the network. The distribution of content and the coordination of content requests is mediated through a centralized server system, which maintains a directory catalog of the available content and of the location of the content within the network.[0028]
• In the preferred embodiments of the present invention, each unit of content, typically represented by a content file, is segmented into discrete parts that are each uniquely identified in the catalog maintained by the mediation server system. Multiple copies of each segment are preferably distributed to cache stores maintained throughout the network to ensure redundant sources of segments for any requesting peer computer system. The distribution of segments within the network of caches is determined by the mediation server system. While entire sets of segments may be distributed to individual peer computer systems, the mediation server system can also operate to ensure that only fragmentary portions of content units are stored by individual peer computer systems. Through fragmentary storage, the effective security of the corresponding content units is fundamentally increased. The redundant distribution of segments permits transfer of the entire set of segments to a requesting peer with an assured high quality of service.[0029]
• The mediation server system preferably manages the transfer of content segments within and between the various storage tiers of content distribution network, including content seeding peer computer systems, dedicated content distribution platforms, and end-user client node computer systems. The seeding peers preferably operate as the source of new content segments for distribution to the content distribution network and as ultimate backup sources for segments of requested content. The dedicated content distribution platforms preferably operate as a middle tier for content distribution, affording a greater fan-out of the transfer load in distributing content segments to the end-user client systems. These content distribution platforms may also be used as dedicated sources of proprietary or other content that for licensing or other reasons will not be distributed for persistent storage in the end-user tier of content caches. Finally, the end-user client node tier is typically a highly heterogenous collection of typically independently operated computer systems, each used to host a segment storage cache and to participate on an ad-hoc basis in the content distribution network. Client node systems may support caches of varying size, network connections of varying capacity, and be available on independent schedules.[0030]
• Requests for selected content units and cache content update requests are submitted to the mediation server through preferably persistent network connections. Manifest lists of the segments may be returned directly or indirectly through an identification of a location within the peer network where a copy of the manifest list is stored. Based on a manifest list, content segments are independently requested by and transferred to nodes of the content distribution system. The peer driven segment retrieval process is cooperatively monitored by the mediation server and, as needed, alternate source locations for segments are provided. Information on the performance of individual peers and the patterns of requests are collected and evaluated on a generally dynamic basis for the generation of request manifest lists. This information is also utilized as a basis for the generation of cache update manifest lists, used to control the background transfer and controlling the storage distribution of content segments throughout the network to optimize the delivery of content in anticipation of demand.[0031]
• A preferred architecture of a[0032]CDN system10, consistent with the present invention, is shown in FIG. 1. TheCDN system10 preferably includes acentral server system12 and a peercontent storage network14. While logically operating as a centralized system, the various server computer systems that cooperatively function as thecentral server system12 can be remotely located, duplicated, and scaled as needed for management, performance, and commercial requirements. The operational functions of thecentral server system12 include the preparation, including segmentation, of new content for publication, the distribution and management of content segments throughout the peercontent storage network14, monitoring the effective performance and actual content segment transfers between various peer nodes within thecontent storage network14, and responding to content and cache update requests.
• In the preferred embodiments of the present invention, new content is initially prepared through a[0033]content publisher system16 by encoding or transcoding a new content file or other unit of content to one of several defined media content formats. The currently preferred formats include the Microsoft® WMA streaming media and the Motion Picture Experts Group MPG3 formats. Other formats can be equivalently processed and used. The content file is then encrypted through a license encoding server. In the preferred embodiments of the present invention, a Microsoft digital rights management (DRM) encryption system is utilized to encrypt the content and subsequently manage the serving of licenses by alicense server20. A content unit identifier, uniquely corresponding to the encrypted, encoded content file, and the DRM generated license key are provided to thelicense server20.
• Encrypted, encoded content files are segmented by a[0034]content segmentation server22. As illustrated in FIG. 2A, representing a first preferred embodiment, acontent file24 is divided intosegments26_1-N, each with a defined segment size, generally within the range of 25 kilobytes and 2 megabytes and typically on the order of 100 kilobytes. Each segment is then assigned the uniquecontent unit identifier28 and asegment sequence identifier30, permitting aparticular content file24 to be reassembled in order from a collection of thecontent segments26_1-N. The resulting construction of named segments32 are then transferred to aseeding peer server34 and stored in a seedingcache36 for subsequent distribution further into the peercontent storage network14. Preferably, theseeding peer server34 is considered part of the peercontent storage network14.
• Segment catalog records[0035]38_1-Nare generated in correspondence with the named segments32_1-N. For a namedsegment322, the correspondingsegment catalog record382 includes a copy of thecontent unit identifier28 andsegment sequence identifier30 of the namedsegment322. Asecurity value40, based on the data content of thesegment262, and afield42 permitting storage of one or more location identifiers are also included in thesegment catalog record382. Preferably, thesecurity value40 is an MD5 hash, multi-byte checksum, or other data value signature of thesegment26₂sufficient to subsequently authenticate the data integrity of thesegment26₂.
• The location identifiers are preferably surrogate client identifiers assigned to the peer nodes within the[0036]content storage network14. These surrogate client identifiers are preferably resolvable by thecentral server system12 to peer network storage cache addresses, preferably in a uniform resource identifier (URI) form. The current URI of a client node can be determined whenever the client node reconnects with thecentral server system12. In resolving a location identifier for a client computer system that is currently unavailable, a null URI is returned. Preferably, location identifiers uniquely correspond to peer nodes. In an alternate embodiment, where the location identifier further identifies a particular cache store of named segments32, the location identifier revolves to a URI identifying a node and cache combination.
• The segment catalog records[0037]38 are stored to a content catalog database maintained by adatabase server44. As created, thelocation field42 of the segment catalog records38 initially contain only the location identifier of the seedingpeer34. Whenever a named segment32 is copied to or deleted from a segment cache within thestorage content network14, thelocation field42 the correspondingsegment catalog record38 is updated to reflect the current set of location identifiers specifying the content caches from which the segment can be obtained.
• In response to a content unit request, the corresponding segment catalog records[0038]38 are prepared and returned as part of a content manifest. In preparing the content manifest, the individual segment catalog records38 are expanded by resolving the location identifiers to the complete URIs for the referenced named segments32. The requesting peer thus receives the necessary information to directly retrieve and validate the named segments32 needed to reconstruct the requestedcontent file24.
• A second preferred embodiment of content segmentation and cataloging is shown in FIG. 2B. As before, named segments[0039]32 are created from thecontent file24. Groups of segments, preferably representing contiguous portions of thecontent file24 are transferred to the seedingpeer34 and subsequently distributed as segment groups to content segment caches within thecontent storage network14.
• A[0040]content manifest file38′ is generated in combination with the named segments32. Thecontent manifest file38′ includes the uniquecontent unit identifier28, alist30′ of the segment sequence identifiers for the named segments32, and the security values40′ for each of the named segments32. In the preferred embodiment of the present invention, thecontent manifest file38′ is distributed as an implied first member of each segment group. Alternately, thecontent manifest file38′ may be distributed separately through the seedingpeer34 to the content segment caches of selected, typically high availability peer computer systems within thecontent storage network14.
• A series of manifest catalog records[0041]38″_Xare also created in combination with the named segments32 andmanifest content file38′. Eachmanifest catalog record38″_Xis established for a respective segment group for thecontent file24. Amanifest catalog record38″_Xincludes the uniquecontent unit identifier28, alist30″ of the segment sequence identifiers for the corresponding sequence group of named segments32, amanifest security value40″, and one ormore location identifiers42″ that specify the content caches storing the corresponding segment group of thecontent file24.
• The[0042]manifest security value40″ is preferably an MD5 hash, multibyte checksum, or other data value signature of thecontent manifest file38′ sufficient to subsequently authenticate the integrity of thecontent manifest file38′. Where the content manifest file is separately distributed, an additional set of one or more location identifiers are included with theidentification identifiers42″ to specify available content segment caches that store copies of thecontent manifest38′. Thelocation identifiers42″ as stored by the content catalog are updated as thecontent manifest38′ and segment groups are copied between and deleted from content segment caches within thecontent storage network14.
• In response to a content unit request, the manifest catalog records[0043]38″_Xare returned to the requesting peer computer system. Thelocation identifiers42″ are expanded URIs prior to returning themanifest catalog record38″. The requesting peer computer system can then obtain a copy of themanifest content file38′ and validate the copy against themanifest security value40″. Individual named content segments32 can then be requested from any peer computer system that persistently stores an encompassing segment group.
• This second preferred embodiment of content segmentation and cataloging is presently preferred based on the reduced load incurred by the[0044]central server system12. Thecontent manifest file38′, due at least to the number and size of includedsecurity values40′, may be an appreciable fraction of the size of the correspondingcontent file24. Distribution and retrieval of thecontent manifest file38′ from thecontent storage network14 greatly reduces the file transfer load imposed on thecentral server system12.
• In the preferred embodiments of the present invention, the[0045]central server system12 supports persistent connections established between the remotely distributed nodes of the peercontent storage network14 and a persistentnetwork proxy server46. The persistent connections are preferably initiated by a client node of the peerconnection storage network14 to a defined TCP/IP socket supported by the persistentnetwork proxy server46. These persistent connections are utilized to permit peer nodes to supply thecentral server system12 with a then current surrogate client identifier, report status and performance information from the peerconnection storage network14 to thecentral server system12 and to request and obtain manifests.
• In the absence of specific activity, as an ongoing background activity, active client nodes utilize the persistent connections to signal a continuing availability to participate in segment data transfers within the peer[0046]content storage network14. In response to activity, client nodes also report outbound segment transfer load levels and other performance indicators affected by ongoing peer network participation, including any communication failures that may occur. Client nodes actively performing inbound segment data transfers from other nodes of the peerconnection storage network14 preferably also utilize the persistent connections to report network data transfer rates and latency information against other identified client nodes. The persistentnetwork proxy server46 preferably collects and records this status and performance information in a network status database maintained by thedatabase server44. The data contained within the network status database, in effect, represents a peer network map useful to plan use of the peercontent storage network14.
• Content unit requests, as submitted by the client nodes, are directed through the persistent[0047]network proxy server46 to ahost broker server48. Each content unit request provides the unique content unit identifier for the requested content unit. A search of the content catalog database locates the catalog records corresponding to the content segments necessary to construct the requested content unit. The referenced locations identifiers are evaluated against the performance information represented by the network peer network map to select an optimal, redundant set of content segments. This evaluation preferably reflects a load-balancing of the ongoing segment data transfer demands on the potentially participating nodes of the peerconnection storage network14. The evaluation also considers the reported network data transfer rates between the requesting client node or similarly situated client nodes and the potentially participating nodes. As a product of the evaluation, thehost broker server48 can produce a content records listing a set of content segments, collectively representing the requested content unit, that can be retrieved from specified locations within the peercontent storage network14. The specified locations determined by thehost broker server48 thus collectively represent a mediated balancing of the need to broadly distribute the system-wide content segment transfer load across the available peer network nodes and ensure effectively uninterrupted delivery of each requested content unit to the requesting peer nodes.
• Preferably, a complete content manifest-based working specification of the requested content unit is dynamically constructed by a requesting client node. The content manifest file is retrieved either from another client node or through the[0048]host broker server48. Based on the segment catalog records38_Xor the segment group index records38″_X, including the content unit and segment sequence identifiers, segment security values, and the expandedlocation identifiers42,42″, the working content manifest is constructed by the client node. Preferably, the expanded location identifiers are presented in a priority ordered by segment sequence number and relatively preferred location from which to transfer the segment as determined by the performance and load-balancing evaluation performed by thehost broker server48.
• The distribution of named content segments[0049]32 throughout the peerconnection storage network14 is, in the preferred embodiments of the present invention, actively performed by the peer nodes of the peerconnection storage network14. Named content segments32 are placed by acontent segmentation server22, individually or as members of segment groups, in thecontent segment cache36 of one or moreseeding peer servers34. The named content segments32 are then dispatched typically through theInternet50 in response to peer node requests to various dedicated contentdistribution network platforms52,client node platforms54,56, and potentiallyother seeding servers34.
• Preferably, all peer nodes within the peer[0050]connection storage network14 implement content servers, such aspeer node applications58,60, to support the network transfer of named content segment32 between respective localcontent segment caches36,62,64. The progressive distribution, including redistribution, of named content segments32 is predominately effected by thepeer node applications58,60 and any secondaryseeding peer servers34 directly requesting sets of named content segments32 for storage in the associatedcontent segment caches36,62,64. Additional distribution and redistribution of named content segments32, again individually or as members of segment groups, follows from the on-demand transfer of the named content segments32 of content units requested by individual client nodes. As named content segments32 are received by a requestingclient node54, a copy is stored at least transiently in the associatedcontent segment cache64 pending streaming to aclient media player66.
• Strategic control over the distribution of the named content segments[0051]32 is preferably performed by amanifest manager server68. In accordance with the present invention, at least the dedicated contentdistribution network platforms52 andclient node platforms54,56 periodically issue cache update manifest requests to thecentral server system12 through the persistent connections with thenetwork proxy server46. Cache update manifest requests are also preferably issued on each initiation of a persistent connection. A content unit request may also be treated as a cache update manifest request.
• In the preferred embodiments of the present invention, the[0052]manifest manager server50 determines a distribution pattern of named content segments32 based on an ongoing analysis of the peer network map, a log of the recent content unit requests and segment transfers, as stored by thebroker server48 to thedatabase server44, and optionally constraints and hints provided bycentral server system12 administrators. In general, the goal of the analysis is to maximize the availability of named content segments32 to likely requesting peer nodes over network connections of sufficient, reliable bandwidth, subject to the segment cache storage size, load limitations, and reported peer node relative network connection bandwidth of individual peer nodes. While, in a present implementation, this dynamic analysis is performed on a progressive batch basis, a near real time evaluation and analysis of the collected data is preferred.
• Constraint information may be employed in the analysis to restrict the distribution of the named content segments[0053]32 of particular content units to defined dedicated contentdistribution network platforms52, as may be externally determined appropriate for certain types of content. Constraint information may also specify language or other meta-data attributes of content units that can be actively considered in the analysis to determine an appropriate distribution pattern for content units. Other constraint information may be provided to specify periods within which specific content units will be available, permitting controlled, progressive distribution of content segments prior to a release date and subsequent retirement after a close date.
• Hinting information is preferably provided to the[0054]manifest manager70 to peremptorily drive the distribution of named content segments32. The hinting information may be specified in terms of the priority and prevalence of the distribution of named content segments32 corresponding to particular content units. The prevalence hints may indicate desired levels of redundant copy distribution over geographic and other domains. Alternately, or in addition, the hinting information may be specified by associating empirical and historically-derived distribution patterns with specified content units. Particularly in the case of historically-derived patterns, the distribution of previously distributed content units can be used as a reference for projecting the likely demand distribution of newly released content units.
• The[0055]central server system12 preferably includes one or more web servers72, which may be geographically distributed, to provide typically end-user accessible interfaces for the selection of content units. The web servers72 connect through network connections to thedatabase server44 to obtain browsable and searchable lists of the content units available through the mediating operation of thecentral server system12. In accordance with the present invention, select web servers72 may be designated as the sole or limited selection source for defined content units. Consequently, these select web servers72 may be operated as the apparent source of proprietary or branded content, at least from the perspective of end-users, yet obtain the full use and benefit of theCDN system10 in distributing the proprietary and branded content units.
• A preferred embodiment of a[0056]peer application60 is detailed in FIG. 3. Executed as a component of a system application on aclient platform54, asystem manager80 implements the top-level procedural logic of thepeer application80. Anetwork connection agent82 provides a persistent proxy interface to thenetwork proxy46, supporting the bidirectional transfer ofcontrol messages84, such as content unit and cache management requests, anddata86, including request and cache update manifests. Named content segments32 are requested and received, in a preferred embodiment of thepeer application60, through anHTTP client component88 from remote peers. Afile receiver component90, supervised by thesystem manager80, performs the detailed transfer control of data files and named content segments32 through theconnection agent82 andHTTP client88 relative to a localcontent segment cache64.
• As each named content segment[0057]32 is received, a security value is regenerated based on the data of containedsegment26 and compared against acorresponding security value40,40′, as provided in the current request or cache update manifest. A comparison failure with thesecurity value40,40′ indicates a corrupt named content segment32, which is discarded. Valid named content segments32 are stored to acontent segment cache64 under the management control of acache manager component92 and thesystem manager80. Preferably, thecontent segment cache64 is encrypted subject to a DRM license. Accesses to the named content segments32 require an encryption key acquired through alicense manager component94, which provides aninterface96 to aconventional DRM client68 and, as required through theconnection agent82, to theremote license server20.
• The[0058]peer application60 preferably implements anHTTP server98 to provide conventional streaming content connectivity to an externalclient media player66 or other streaming media content client. A streamingcontent component100 coordinates between thesystem manager80, for initial set-up of the content streaming session, and thecache manager92, for the ordered retrieval of named content segments32 corresponding to a requested content unit. Retrieved named content segments32 are progressively passed by the streamingserver100 from the localcontent segment cache64 to theHTTP server98 for relay to amedia player66.
• The[0059]HTTP server98 also supports named content segment32 transfer requests from other peer nodes of the peerconnection storage network14. Asegment server component102 is utilized to manage named content segment32 transfers, subject to segment transfer session management performed by thesystem manager80. The transfer of named content segments32 to theHTTP server98 is coordinated by thesegment server100 with thecache manager92 for selection of the request identified named content segments32 from thecontent segment cache64.
• A preferred[0060]architecture110 of a content delivery networkcentral server system12, exclusive of segment preparation and publication components, is shown in FIG. 4. A conventional hardware-based networkconnection load balancer112 supports a scalable set ofCDN server systems114,116. EachCDN server systems114,116 implements a set of executable server components that are implemented on one or more conventional network connected server computer systems. TheCDN server systems114,116 share access to aCDN database118 configured to store OLTP accessible data and an archive database providing storage of recently logged and historical data that can be used for analytic and reporting purposes.
• A[0061]client proxy component120 maintains the direct socket connections for the persistent client sessions established against the active peer nodes of the peerconnection storage network14. A startup message is received by theclient proxy component120 from each peer node upon joining the peerconnection storage network14. Completion messages are received as different processes are completed by the peer nodes.
• A[0062]proxy manager122 monitors the connections established with theclient proxy component120 to maintain a data structure representing the peer nodes that are currently active and accessible. Periodic status messages are exchanged to actively monitor the state of the connected peer nodes. Failures in the status exchange are preferably analyzed with the result of redirecting a peer node to anotherCDN server system114,116, which may be able to establish a more reliable network connection, or the connection is disconnected and the peer node is identified as inactive.
• A[0063]client session component124 establishes defined contexts for communications with each of the peer nodes connected to theclient proxy120. Within each context, information is gathered through various progress, status, and logging messages received from the peer nodes. The collected information, as well as the activity state information managed by the proxy manager, is stored to theCDN database118 for subsequent use in performance analysis and activity reporting.
• A[0064]host broker component126 receives, through theclient proxy120, the peer node content unit requests. Through OLTP database accesses, the host broker determines an optimal set of peer nodes from which the requesting peer node can download the named content segments32 corresponding to the requested content unit. Preferably, the host broker selects all active peer nodes that store named content segments32, individually or in segment groups, of the requested content unit and then orders identical copies of the named content segments32 by the load level of the source peer node and the evaluated connection speed between the source and requesting peer nodes. The topsegment catalog record38,38″ entries for named content segments32 are selected and provided in one or more request catalog messages that are then returned to the requesting peer.
• A[0065]file session component128 actively monitors the ongoing named content segment32 download and streaming operations of the individual peer nodes within corresponding client sessions. In conjunction with thehost broker component126, a unique file session identifier is provided in each content unit request manifest. Unique file session identifiers are also provided in each cache update manifest. When a streaming content unit transfer is terminated, the peer node provides a session finished message to thefile session component128. A file session finished message is also provided when a peer node has completed a content segment cache update, based on a provided cache update manifest.
• A file session finished message includes the request or cache update manifest corresponding file session identifier and information detailing the required transfer time, number of source peer nodes used, and other statistically relevant information. Network communications failures with particular source peer nodes and other error conditions are also reported. The acquired information is stored to the[0066]CDN database118 for subsequent use in performance analysis and activity reporting.
• A[0067]cache manifest manager130 is responsible for establishing the distribution of named content segments throughout the peerconnection storage network14. The collected information stored by theCDN database118 is periodically evaluated on a daily or shorter basis. Newly available content units, as represented by stored segment catalog records38, are considered in the evaluation. An updated distribution plan is ultimately produced and stored by thecache manifest manager130 to theCDN database118.
• A peer[0068]cache update manager132 is responsive to cache update manifest requests, as periodically issued by the peer nodes. Based on the segment distribution plan determined by thecache manifest manager130 and preferably further qualified by recognition of the ongoing named content segment32 transfers dynamically reported to thefile session component128, a cache update manifest specific to the requesting peer node is generated and returned.
• A[0069]license manager component134 is responsive to license request messages issued by client nodes in connection with content unit requests. The request manifest, in addition to providing a requisite set of segment catalog records38, preferably identifies the type of licensing encryption, if any, applied to the requested content unit. A license request message includes the request correspondingcontent unit identifier28, a license location identifier, which specifies the licensing authority for the requested content and requesting client node, and the license type identifier.
• The specified license type permits the[0070]license manager134 to validate the requested content unit license against the typically external licensing authority. In the preferred embodiments of the present invention, thelicense manager134 utilizes a conventional Microsoft digital rights manager. Where validated, a license key is generated by thelicense manager134 and provided to alicense server136. The license is thus available to theclient media player66 for use in decrypting the streaming media content unit as received through theclient peer application60. A license response message is also returned through theclient proxy120 to the requesting peer node in response to the license request message. The license response message either acknowledges the availability of the license key or provides a validation failure explanation.
• The[0071]preferred process140 implemented by aclient peer application60 is shown in FIG. 5. In connection with the execution of thepeer application60, aclient media player66, Web browser or other client application, executed on theclient platform54, permits an end-user to select and login142 to a chosen Web server72. Preferably, a list of available content units is displayed forselection144 by the end-user. Based typically on an end-user selection, a content unit request is issued146 to theCDN server system114 currently supporting the persistent connection to thepeer application60. The content unit request is brokered148 and arequest manifest150 is returned.
• Upon receipt of the request manifest, the[0072]client peer application60 determines whether an encryption license applies to the requested content unit. Alicense validation154 is obtained where required. The segment catalog records38,38″ provided by the request manifest are parsed and corresponding named content segment32 transfer requests are progressively issued156 to thesegment catalog record38,38″ identified peer nodes. As the requested named content segments32 are received158, the integrity of each named content segment32 is checked160. Valid named content segments are preferably at least transiently stored162 to thecontent segment cache64. An updated cache update manifest provided in combination with the request manifest can determine which named content segments32 are to be persistently retained in thecontent segment cache64. Named content segments32 that fail the integrity check are re-requested from the same or an alternate peer node.
• In the preferred embodiments of the present invention, once at least the initial named content segment[0073]32 has been received, streaming164 of the requested content unit is enabled to the attachedclient media player66. As permitted by theclient media player66, Web browser or other client application, a new content unit can be selected144 at any time, terminating the current transfer, and causing a new content unit request to be issued146.
• Periodic updates of the[0074]content segment cache64 are scheduled by theclient peer application60. Cache update requests are preferably issued166 automatically by thepeer application60 to the currently connectedCDN server system114. A cache update manifest is generated168 and returned170 to the requestingclient peer application60. The cache update manifest is parsed by theclient peer application60 to identify any named content segments32, as specified by corresponding segment catalog records38, that are not currently stored by thecontent segment cache64. Requests for the non-resident named content segments are issued156 and the named content segments32 are received158 and stored162 to thecontent segment cache64.
• In an alternate embodiment of the present invention, the cache update manifest provides meta-information that is used by the[0075]client peer application60 to qualify cache update operations. The cache update manifest meta-information is utilized to specify the schedule of cache update requests and the location of theCDN server system114,116 to use as the target of the next cache update request. The meta-information may also be provided to specify a delay schedule for issuance of named content segment32 transfer requests. This allows thecache update manager132 to fully mediate the data transfer load on the peercontent storage network14 both in terms of selecting the transfer source peer nodes utilized and the temporal distribution of the load imposed on those nodes. Such mediation is particularly valuable to optimally schedule the load placed on the seeding peers34 anddedicated CDN platforms52 particularly where the peercontent storage network14 includes a large number of peer nodes.
• The managed[0076]client peer process180 of named content segment request and retrieval is shown in greater detail in FIG. 6. The contents of request and cache update manifests, including meta-information, are initially parsed182 upon receipt of the manifests. Deferred operations are preferably handled through a periodic re-parsing of the manifests at the deferred time intervals. In anticipation of the receipt of new named content segments32 beyond aclient platform54 defined cache size, named content segments32 no longer identified in the current cache update manifest are deleted184 from thecontent segment cache64.
• Preferably, multiple named content segments[0077]32 are requested156 concurrently from the peercontent storage network14. Each concurrently requested named content segment32 is also redundantly requested from multiple peer node locations. The total number of concurrent named content segment32 transfers allowed is a dependent on the maximum acceptable load permitted on theclient platform54. Excluding the redundant transfers, a default limit is set at four concurrent transfers of unique named content segments32. As redundant copies of named content segments are received158, the transfer bandwidths of each are monitored. Once a reasonably stable gauge of the transfer bandwidths can be determined, adjusted potentially for different transfer start times and the anticipated remaining length of the named content segment32, only the highest bandwidth transfer for each named content segment32 is maintained. A failure to complete any of these remaining transfers is detected186. A severe reduction in the transfer bandwidth is also preferably treated as a transfer failure. Redundant requests for the same named content segment32 are again reissued188 to multiple peer node locations determinable from the manifests.
• Named content segments[0078]32 are stored162 to thecontent segment cache64 as received158. A security value for the segment data may be accumulated as the named content segment32 is received or computed once the named content segment32 once the transfer is completed. This actual security value is then compared190 to thesecurity value40 provided in the correspondingsegment catalog record38. On a comparison failure, the received named content segment32 is deleted from thecontent segment cache64. Redundant requests for the named content segment32 are again reissued188.
• Preferably, detailed information, including the connection latency, average bandwidth and reliability of transferring named content segments[0079]32 relative to the requestingclient platform54, is collected by theclient peer application60. Information detailing transfer failures and data integrity failures, along with the identity of the peer nodes participating in the failed transactions, is also collected. This performance information is reported to the connectedCDN server system114,116, preferably in connection with the transfer completion of each content unit transfer and cache update.
• Thus, an efficient peer-to-peer content distribution network system architecture capable of efficiently providing a high quality of service in the delivery of multimedia data streams to end-users has been described. While the present invention has been described particularly with reference to operation over the public Internet, the present invention is equally applicable to the distribution over other public and private communications networks. Additionally, the present invention is also applicable to the rapid and efficient distribution of digital information that may have use other than as streaming media content.[0080]
• In view of the above description of the preferred embodiments of the present invention, many modifications and variations of the disclosed embodiments will be readily appreciated by those of skill in the art. It is therefore to be understood that, within the scope of the appended claims, the invention may be practiced otherwise than as specifically described above.[0081]

Claims (19)

1. A content distribution system supporting the transfer of a content file present on a plurality of peer computers to a recipient computer, said content distribution network comprising:

a) a content catalog containing an identification of a content file and a plurality of location identifiers, wherein each location identifier includes an identification of a peer computer and of a named subunit of said content file; and

b) a control processor, coupled to said content catalog and coupleable through a network connection to a recipient computer, wherein said control processor, responsive to a request identifying said content file provided by said recipient computer, provides a content file manifest to said recipient computer, said content file manifest including said plurality of location identifiers;

whereby said recipient computer can retrieve and assemble corresponding named subunits of said content file from said peer computers to obtain said content file.

2. The content distribution system ofclaim 1 further comprising a performance database, coupled to said control processor, containing performance information for said peer computers, wherein said control processor provides said content file manifest with an ordered listing of said plurality of location identifiers, and wherein said control processor determines said ordered listing based on said performance information.

3. The content distribution system ofclaim 2 wherein said content file manifest provides multiple location identifiers for said named subunits and wherein said ordered listing provides a preference order for the retrieval of said named subunits by said recipient computer.

4. The content distribution system ofclaim 3 wherein said control processor is further coupleable to said recipient computer to receive update performance information, wherein said control processor provides for the storage of said update performance information in said performance database.

5. The content distribution system ofclaim 4 wherein said update performance information includes activity information with respect to said recipient computer and a predetermined one of said peer computers.

6. A content distribution system comprising:

a) a peer computer network including a plurality of segment data caches that provide for the distributed storage of named content segments, wherein said plurality of segment data caches have associated cache location identifiers, said peer computer network being responsive to named segment requests including named segment identifiers to provide corresponding named content segments;

b) a content broker computer system including a content catalog that provides for the storage of a named content identifier in association with a plurality of named content segment identifiers further associated with a plurality of cache location identifiers, said content broker computer system being responsive to a broker request including said named content identifier to provide said plurality of cache location identifiers selectively ordered with respect to said plurality of named content segment identifiers; and

c) a request computer system coupleable through a communications network to said content broker computer system to provide said broker request and to said peer computer network to provide segment requests for said plurality of named content segments respectively associated with selected cache location identifiers determined based on the selective ordering of said plurality of cache location identifiers with respect to said plurality of named content segment identifiers.

7. The content distribution system ofclaim 6 wherein said request computer system is operative to provide segment requests to said peer computer network to progressively obtain said named content segments corresponding to said named content identifier.

8. The content distribution system ofclaim 7 wherein said content broker provides said request computer system with a manifest presenting preference orders of cache location identifiers for respective ones of said plurality of named content segments.

9. A method of operating a content distribution network to support the retrieval of a streaming content file, said method comprising the steps of:

a) distributing a plurality of named content segments, collectively representing a predetermined streaming content file, among a plurality of peer caches, wherein a portion of said named content segments are multiply distributed among said plurality of peer caches; and

b) responding to a request for said predetermined streaming content file by providing a retrieval list identifying said plurality of named content segments, said list including an identification of the distribution of said plurality of named content segments;

whereby said plurality of named content segments can be retrieved from among said plurality of peer caches to obtain said streaming content file.

10. The method ofclaim 9 further comprising the step of generating said retrieval list with a plurality of location lists, wherein each said location list identifies, with respect to a predetermined named content segment, a subset of said plurality of peer caches to which said predetermined named content segment was distributed.

11. The method ofclaim 10 wherein said step of generating provides for an ordering of the identification of said subsets of said plurality of peer caches.

12. The method ofclaim 11 further comprising the step of collecting performance information with respect to said plurality of peer caches and wherein said step of generating provides for said ordering based on said performance information.

13. The method ofclaim 12 further comprising the step of cataloging the locations of said plurality of named content segments as distributed among said peer caches.

14. The method ofclaim 13 further comprising the step of segmenting said predetermined streaming content file into said plurality of named content segments.

15. A content distribution system supporting the delivery of streaming media content files over a communications network to a recipient computer system, said content distribution system comprising:

a) a catalog of predetermined content files, said catalog including a listing of content file segments, wherein predetermined pluralities of said content file segments correspond to said predetermined content files;

b) a predictive content distribution system operative to distribute copies, including multiple copies, of said content file segments to a plurality of peer computer systems, said predictive content distribution system being coupled to said catalog to associate peer system identifiers with said content file segments to reflect the distribution of said content file segments to said peer computer systems;

c) a peer computer management system coupleable to said plurality of peer computer systems and operative to monitor the availability of said content file segments as distributed to said plurality of peer computer systems; and

d) a request broker system coupleable through said communications network to receive a content file request from a recipient computer system, said request broker system providing, based on the availability of said content file segments as determined by said peer computer management system, a manifest of content file segment identifiers and associated peer system identifiers;

whereby said recipient computer system can utilize said manifest to obtain the content file segments corresponding to the content file identified by said content file request.

16. The content distribution system ofclaim 16 wherein said peer computer management system dynamically monitors said plurality of peer computer systems.

17. The content distribution system ofclaim 17 wherein said request broker system dynamically generates said manifest in response to said content file request.

18. The content distribution system ofclaim 18 wherein said peer computer management system maintains activity and capability information with respect to said plurality of peer computer systems and wherein request broker system is responsive to said activity and capability information in generating said manifest.

19. A method of obtaining streaming media content files over a communications network, said method comprising the steps of:

a) issuing, over a communications network, a content request for a predetermined content file to a host computer system;

b) receiving a manifest from said host computer system, said manifest including a list of content segments corresponding to said predetermined content file, said list including a peer cache location identifier for each said content segment; and

c) requesting, over said communications network, a first plurality of said content segments from a second plurality of peer caches identified by said peer cache location identifiers corresponding to said first plurality.

Images