JP2014042124A - Content distribution system and content distribution method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve cache efficiency in a cache server of a content distribution system.SOLUTION: An origin server 10 measures a viewing separation characteristic of respective contents and distributes a chunk of the contents until a viewing separation rate becomes α to respective cache servers 20. Subsequent data gives a request to the origin server 10 from the cache server 20 only when a user terminal 30 gives the request. When idle capacity for storing the new content lacks, the chunk of the contents whose possibility to be viewed from the user terminal 30 is low is preferentially deleted by referring to reproduction time until the number of accesses of the respective contents, the number of reproduction completions and a separation rate become β.

Description

  The present invention relates to a content distribution system and a content distribution method, and particularly to a video content cache management technique in a cache server on a content distribution network.

  In recent years, traffic due to video content (moving content) has increased rapidly on the Internet. As such video content distribution technology, CDN (Content Delivery Network) and streaming of video content using HTTP (HyperText Transfer Protocol) as disclosed in Non-Patent Document 1 are used. In CDN, in order to deliver content quickly and efficiently, the content of the origin server is copied to cache servers distributed on the network, and the content requested from the user terminal is held in the cache server Delivers content from a cache server instead of an origin server. By distributing from a cache server, CDN can reduce the load on the origin server, reduce network equipment costs by reducing traffic, and improve QoE (Quality of Experience) by reducing response time. In order to reduce the traffic between the origin server and the cache server at the peak time and to reduce the load on the origin server, it is effective to pre-distribute the content to the cache server at the off-peak time.

R. Pantos, Ed., W. May, “HTTP Live Streaming” IETF Internet-Draft, March 2012.

  In particular, video content that is highly likely to be large-capacity data is highly effective in reducing traffic when a cache hit occurs (content requested from the user terminal is held in the cache server). On the other hand, video content has characteristics such as a long tail of popularity and short-term fluctuations in popularity, making it difficult to predict popularity. Therefore, in order to improve the cache hit rate, it is necessary to store a large number of video contents in the cache server cache, but the cache capacity per cache server is limited.

  Conventionally, since the content pre-cache generally holds the entire content data in the cache, the number of contents that can be held in the cache is limited. Therefore, the cache hit rate of the cache server is reduced. Furthermore, in a service that allows users to watch movies and dramas for free or for a fixed amount that has been widely used in recent years, there are frequent situations in which a user stops watching only the first video content. In this case, the data after the viewing and leaving time is not used among the entire data of the corresponding content held in the cache. Especially for long-time video content such as movies, the proportion of data that is not viewed becomes very large, and the cache efficiency of the cache server (the proportion of content data held in the cache that is actually viewed by the user terminal) There is a problem that becomes low. Accordingly, an object of the present invention is to solve the above-described problems and improve cache efficiency in a cache server of a content distribution system.

In order to solve the above-described problem, the present invention is a content distribution system including a cache server that distributes video content in response to a request from a user terminal, and an origin server that distributes the content to the cache server. The origin server includes a content information table indicating the content, a content group to which the content belongs for each content, and a viewing leaving rate defined by the following formula (1) for each content group Gx of the content: An origin server storage unit that stores a viewing detachment characteristic table indicating the playback time T (Gx, α) of the content of the content group Gx having a predetermined value α, and the viewing detachment rate at the content playback time t seconds = (Number of views at the start of playback of the content-the content Number of viewings at the playback time t seconds) / (number of viewings at the start of playback of the content) Formula (1)
Referring to the viewing / leaving characteristic table and the content information table, content that distributes data from the beginning until the playback time T (Gx, α) is exceeded to the cache server among the content data of the content group Gx A cache server storage unit for storing the distributed content data, and the content data to the user terminal that is the transmission source of the request in response to a request from the user terminal. A content delivery unit for delivery; and an origin server communication unit for sending a delivery request for the content data to the origin server when the content data requested from the user terminal is not in the cache server storage unit.

  By doing in this way, the data of a part of content with high possibility that a user terminal will view is delivered to each cache server. Thereby, the cache efficiency in the cache server can be improved. It should be noted that the access count of the content is used as the “viewing count at the start of content playback” in the equation (1). In addition, the “number of times content is viewed in playback time t seconds” uses the number of times content is viewed (played back) in playback time t seconds.

  Further, the cache server storage unit of the content distribution system of the present invention further stores a content management table, and the cache server further stores the data size of the content data to be distributed included in the inquiry from the origin server A cache management unit for determining whether or not the cache server storage unit has a free capacity for caching the content data from the origin server, and distribution of the content to the user terminal for each content A content reproduction recording unit that records the number of times and the number of completions of reproduction in the user terminal in the content management table, and the origin server communication unit further transfers the content data from the origin server to the cache server storage unit. Cash When the inquiry about whether or not there is a free capacity to be received is received, the data size of the content data to be distributed included in the inquiry is output to the cache management unit, and the content data is cached by the cache management unit When it is determined that there is free space, the content data distribution request is transmitted to the origin server, and the cache management unit determines that there is no free space to cache the content data from the origin server , Referring to the content management table as data to be deleted, preferentially selecting from the content that is already stored in the cache server storage unit, the content that has a small number of distributions and a small number of reproductions, Until the free space can be secured, The process is repeated to remove a-option content.

Further, the content reproduction recording unit in the cache server of the content distribution system of the present invention further records, in the cache chunk information table, what data of the content has been reproduced when the reproduction of the content is stopped, The cache server storage unit further stores the cache chunk information table. When the cache management unit determines that there is no free capacity to cache the content data from the origin server, the content is deleted as the data to be deleted. Referring to the management table and the cache chunk information table, even if the content with the distribution count of 0 is searched from the content already stored in the cache server storage unit, and the content with the distribution count of 0 is deleted, Free space is secured If not, the content is selected preferentially from the content with the smaller number of distributions and the number of completed reproductions, and among the data of the selected content, the departure rate from the reproduction start defined by the following equation (2) is The data after the time exceeding the predetermined threshold β is obtained, and the process of deleting the obtained data is repeated until the free space can be secured.
Exit rate of the content at the playback time t seconds = Number of playback cancellations from the start of playback of the content to the playback time t seconds / Number of times the content is distributed. Formula (2)

  In this way, each cache server deletes data that is unlikely to be viewed from the content data held by itself when the free capacity of the cache server storage unit is insufficient. As a result, the cache server holds the content data of the portion that is more likely to be viewed by the user terminal, so that the cache efficiency of the cache server can be improved.

  In addition, the content transmission unit and the content distribution unit in the content distribution system of the present invention distribute the data of the content in a chunk format.

  In this way, since chunks are used in data transmission of content from the origin server to the cache server and data distribution from the cache server to the user terminal, up to what data has been distributed to the cache server and user terminal, and subsequent data When distributing, it becomes easier to manage which data should be distributed.

  ADVANTAGE OF THE INVENTION According to this invention, the cache efficiency of a cache server can be improved in the content delivery system which delivers a content from an origin server to a cache server in advance.

It is a figure explaining the outline | summary of the content delivery system of this Embodiment. It is the figure which illustrated the viewing-and-listening characteristic for every content group. It is a block block diagram of the origin server of FIG. It is a block block diagram of the cache server of FIG. It is the figure which showed the processing procedure (procedure which an origin server delivers a content (chunk of content) to a cache server) of the content delivery system of FIG. It is the figure which showed the processing procedure (The processing procedure of the cache server which received the content delivery request from a user terminal) of the content delivery system of FIG. It is the figure which showed the processing procedure (processing procedure when a user terminal stops reproduction | regeneration of a content) of the content delivery system of FIG. It is the figure which showed the processing procedure (chunk deletion procedure by a cache server) of the content delivery system of FIG.

<Overview>
Hereinafter, embodiments of the present invention will be described with reference to the drawings. First, an outline of the content distribution system of the present embodiment will be described with reference to FIG. Note that the content handled in this embodiment is mainly video content. In addition, the content is transmitted and distributed in units of chunks, but may be data cut at an arbitrary location of the content.

  As shown in FIG. 1, the content distribution system includes an origin server 10 that distributes content to the cache server 20 in advance, and a cache server 20 that distributes content in response to a request from the user terminal 30. The number of each server and user terminal 30 is not limited to the number shown in FIG. The origin server 10 and the cache server 20 are linked and may be equipped with functions such as TCP (Transmission Control Protocol) speedup and file compression for high-speed data transfer between the two servers.

  The origin server 10 obtains the viewing detachment characteristic for each content group from the viewing detachment rate of each content measured in advance (defined by equation (1) described later). A content group is a group of content groups having the same content attribute. An example of the viewing withdrawal characteristic is shown in FIG. The viewing withdrawal rate at the playback time t seconds is defined by the following equation (1). Here, the number of views is the number of times the user terminal 30 reproduces (views) the content.

  Viewing disconnection rate at time t of playback of content = (number of views at the start of playback of the content−number of views at the time of playback of the content at t seconds) / (number of views at the start of playback of the content) 1)

  From this viewing leaving characteristic, for each content group Gx (x = 0, 1,...), A playback time T (Gx, α) at which the viewing leaving rate becomes a predetermined value α (α is appropriately set, for example, 0.9). ) For example, the playback time when the viewing / leaving rate of the content group of the content group G0 is α is a value indicated by T (G0, α) in the graph of FIG. 2, and the viewing / leaving rate of the content group of the content group G1 is α. Is a value indicated by T (G1, α) in the graph of FIG.

  Then, the origin server 10 distributes the content data (chunk) up to T (Gx, α) to the cache server 20 in advance for the content group of the content group Gx.

  For example, when the chunks up to T (Gx, α) are up to chunk number 3 for the content of the content group Gx, the origin server 10 chunks up to T (Gx, α) (that is, chunk numbers 1 to 3). Are delivered to the cache server 20 in advance. In this way, each cache server 20 holds chunks up to the time when there is a high possibility of being viewed by the user terminal 30 among the chunks constituting the content. Efficiency can be improved. The subsequent data of the pre-distributed data (chunk) is requested from the cache server 20 to the origin server 10 when requested from the user terminal 30, and is distributed to the user terminal 30.

  Here, the cache server 20 records the number of accesses, the number of completed reproductions, and the detachment rate for each chunk held by itself. Here, the separation rate is defined by the following equation (2).

  Exit rate = sum of the number of playback stops from the start of playback of the content to the chunk / number of accesses to the content (number of distributions) (3)

  When content data is received from the origin server 10 and the cache server 20 does not have enough free space to receive the data, among the contents held by the cache server 20, Delete content (content chunks) based on the number of completed playbacks and withdrawal rate. For example, the cache server 20 deletes content with a relatively small number of accesses and content with a relatively small number of playback completions, or deletes data of chunks ahead of a chunk whose withdrawal rate is β (the value of β is set appropriately) To do. For example, in the cache server 20, when the chunk whose content detachment rate is β is the chunk with the chunk number 2 illustrated in FIG. 1, the chunks after the chunk number 3 of the cache server 20 are deleted. The cache server 20 receives content data from the origin server 10. In this way, the cache efficiency of the cache server 20 can be further improved.

<Origin server>
Next, the origin server 10 will be described in detail with reference to FIG. The origin server 10 includes an origin server storage unit 15, a pre-cache target selection unit 11, and a content transmission unit 12. The viewing detachment characteristic table creation unit 13 indicated by a broken line may or may not be equipped, and the case where it is equipped will be described later.

  The pre-cache target selection unit 11 and the content transmission unit 12 are realized by a program execution process by a CPU (Central Processing Unit) included in the origin server 10, a dedicated circuit, or the like. Further, the origin server storage unit 15 is realized by a storage medium such as a random access memory (RAM), a read only memory (ROM), a hard disk drive (HDD), and a flash memory. When the origin server 10 is realized by program execution processing, the origin server storage unit 15 stores a program for realizing the function of the origin server 10.

  The origin server storage unit 15 includes a content 151, a viewing / leaving characteristic table 152, a chunk information table 153, and a content information table 154.

  The content 151 is content such as a moving image, and is divided into a plurality of chunks in advance.

  The viewing detachment characteristic table (viewing detachment characteristic information) 152 is a reproduction time T (Gx,) for each content group in which the viewing detachment rate for each content group obtained based on the viewing detachment rate of the content belonging to the content group is α. α) (see Table 1). This information is created based on the content viewing / listening rate of each content group (see FIG. 2 described above).

  It should be noted that the content group G2 in FIG. 2 has a viewing leaving rate smaller than α even when the playback time exceeds a predetermined value (not shown), that is, a content group that is predicted to be viewed for a long time. Thus, the reproduction time T (Gx, α) = 0.

  The chunk information table 153 shows information regarding each chunk constituting the content for each content of the content group. For example, as shown in Table 2, the chunk information table 153 includes, for each content, the chunk number of the chunk that constitutes the content, the playback time length of the chunk of the chunk number, the data size, the URI (Uniform Resource Identifier), etc. Is stored. In the chunk information table 153, chunk information is stored in chronological order from the first chunk constituting the content.

  The content information table 154 is information indicating, for each content, the content name (content identification information) of the content, the total number of chunks constituting the content, and the content group to which the content belongs.

  The prior cache target selection unit 11 refers to the viewing / leaving characteristic table 152 and selects a chunk of content to be distributed in advance to each cache server 20. Here, the pre-cache target selection unit 11 refers to the viewing / leaving characteristic table 152, the chunk information table 153, and the content information table 154, and the sum of the playback time lengths from the top of the data of each content is T (Gx, Chunks exceeding α) are selected as chunks to be distributed in advance to the cache server 20. Note that the pre-cache target selection unit 11 selects all chunks for content in a content group with a playback time T (Gx, α) = 0 (that is, a content group whose viewing departure rate is smaller than α even if the playback time exceeds a predetermined value). Select as a pre-delivered chunk.

  The content transmission unit 12 distributes the chunk of content (pre-cache target chunk) selected by the pre-cache target selection unit 11 to the cache server 20. In addition, the content transmission unit 12 reads the content chunk requested from the cache server 20 from the origin server storage unit 15 and distributes it. Note that the content transmission unit 12 inquires of the cache server 20 whether or not there is a free capacity for caching the chunk when distributing the chunk of the content. When the content transmission unit 12 receives a response from the cache server 20 that there is free space, the content transmission unit 12 transmits a chunk of content. At this time, the content transmission unit 12 transmits the chunk information and the content information of the chunk to the cache server 20 together with the chunk of the content. The chunk information is obtained by reading out information (for example, chunk number, playback time length, data size, URI, etc.) related to the chunk of the content to be distributed among the information stored in the chunk information table 153. Further, the content information is obtained by reading out the content information (for example, content identification information, the total number of chunks, etc.) of the distribution target chunk from the information stored in the content information table 154.

<Cache server>
Next, the cache server 20 will be described in detail with reference to FIG. The cache server 20 includes a cache server storage unit 21, an origin server communication unit 22, a cache management unit 23, a request reception unit 24, a content distribution unit 25, and a content reproduction / recording unit 26.

  The origin server communication unit 22, the cache management unit 23, the request reception unit 24, the content distribution unit 25, and the content reproduction / recording unit 26 are a program execution process performed by a CPU provided in the communication interface and the cache server 20, Realized by a dedicated circuit or the like. Further, the cache server storage unit 21 is realized by a storage medium such as a RAM, a ROM, an HDD, and a flash memory. When the cache server 20 is realized by program execution processing, the cache server storage unit 21 stores a program for realizing the function of the cache server 20.

  The cache server storage unit 21 includes an area for storing a content chunk (content chunk) 211 transmitted from the origin server 10, a cache chunk information table 212, and a content management table 213.

  As described above, the content chunk 211 is a chunk of content distributed from the origin server 10.

  The cache chunk information table 212 indicates, for each chunk of the content chunk 211, the chunk number, playback time length, data size, URI, cache storage destination (storage destination in the cache server storage unit 21), and the number of playback stops. Information (see Table 4). Information shown in the chunk information received from the origin server 10 is used as the chunk number, playback time length, data size, and URI in the cache chunk information table 212.

  The content management table 213 includes, for each content in the content chunk 211, a content name (content identification information) of the content, the total number of chunks of the content, and a chunk held in the cache server storage unit 21 among all the chunks of the content. Number (the number of retained chunks), the date and time when the content is received from the origin server 10, the number of accesses to the content (the number of distributions to the user terminal 30), and the number of playback completions of the content (the user terminal 30 is the last of the content) The number of times received until the first chunk) (see Table 5). Information shown in the content information received from the origin server 10 is used for the total number of chunks in the content management table 213.

  The origin server communication unit 22 communicates with the origin server 10. The origin server communication unit 22 receives a chunk of content, chunk information of the chunk, and content information from the origin server 10. The received chunk of content is stored in a predetermined area of the cache server storage unit 21, the received chunk information is stored in the cache chunk information table 212, and the received content information is stored in the content management table 213. At this time, the origin server communication unit 22 records the cache storage destination of the received chunk in the cache chunk information table 212 and records the content reception date and time in the content management table 213. The number of playback cancellations in the cache chunk information table 212, the number of accesses and the number of playback completions in the content management table 213 are all set to initial values (0).

  In addition, the origin server communication unit 22 transmits a delivery request for the subsequent chunk of this content to the origin server 10 when the content chunk to be delivered to the user terminal 30 is likely to be insufficient. For example, the origin server communication unit 22 first receives a chunk (pre-cache target chunk) from the origin server 10 and stores the chunk in the cache server storage unit 21 via the cache management unit 23 (described later in detail). Thereafter, when the origin server communication unit 22 distributes the chunk of content to the user terminal 30, the number of subsequent chunks stored in the cache server storage unit 21 is equal to or less than a predetermined threshold (subsequent chunk request threshold). The subsequent chunk delivery request is transmitted to the origin server 10. Then, the origin server communication unit 22 stores the subsequent chunk received from the origin server 10 in the cache server storage unit 21 via the cache management unit 23, and prepares for the distribution of the subsequent chunk from the user terminal 30. The origin server communication unit 22 receives a notification of content chunk delivery from the origin server 10 prior to delivery of the content chunk. At this time, the origin server communication unit 22 outputs the data size of the chunk of content from the origin server 10 to the cache management unit 23, and caches (stores) this chunk in the cache server storage unit 21. Ask if there is any. When the cache management unit 23 determines that there is a free capacity for caching the content chunk, the origin server communication unit 22 requests the origin server 10 to distribute the content chunk.

  The cache management unit 23 reads, writes, and deletes data in the content chunk 211, the content management table 213, and the cache chunk information table 212 in the cache server storage unit 21. When receiving the content chunk from the origin server 10, the cache management unit 23 determines whether or not the cache server storage unit 21 has a free capacity for caching the content data. Here, when it is determined that there is no free space, a part of the content chunk 211 is deleted. In other words, the cache management unit 23 refers to the content management table 213 and the cache chunk information table 212, and prioritizes the content with the smaller number of accesses (number of distributions) and the number of completed reproductions as the contents to be deleted. In the selected content chunk, a portion of the chunk after the time when the departure rate from the start of reproduction exceeds a predetermined threshold β is obtained. Then, the cache management unit 23 reserves a predetermined amount of chunks after the time when the withdrawal rate exceeds the predetermined threshold β (a cache capacity for caching the chunk of content from the origin server 10). delete. Then, the cache management unit 23 deletes and corrects information regarding the deleted content (content chunk) from the cache chunk information table 212 (see Table 4) and the content management table 213 (see Table 5). Details of the chunk deletion procedure will be described later. In addition, the cache management unit 23 clears the number of accesses and the number of completed reproductions in the content management table 213 and the number of reproduction cancellations in the cache chunk information table 212 every predetermined period. This is for preventing the cache management unit 23 from infinitely increasing the number of content accesses and the number of completed reproductions and the number of chunk reproduction cancellations.

  The request receiving unit 24 receives a content (content chunk) distribution request from the user terminal 30. Then, the request reception unit 24 checks whether or not the content chunk 211 has a chunk of the content to be requested, and if so, instructs the content distribution unit 25 to distribute the content to be requested. On the other hand, when there is no chunk of the request target content, the request receiving unit 24 instructs the origin server communication unit 22 to acquire the request target content from the origin server 10. Then, the content chunk acquired from the origin server 10 is added to the content chunk 211, and is distributed by the content distribution unit 25 to the user terminal 30 that is the transmission source of the request.

  The content distribution unit 25 distributes the chunk of content to the user terminal 30 that is the transmission source of the content distribution request.

  The content reproduction recording unit 26 records, for each content, the number of distributions of the content to the user terminal 30 (the number of accesses of the content by the user terminal 30) and the number of reproduction completions at the user terminal 30 in the content management table 213. Then, when the reproduction of the content is stopped, the content reproduction recording unit 26 records in the cache chunk information table 212 what data (chunk) of the content has been reproduced. For example, when it is confirmed that the user terminal 30 has started receiving the distribution of the content k, the content reproduction recording unit 26 adds 1 to the number of accesses to the content k in the content management table 213. In addition, when the content reproduction recording unit 26 confirms that the user terminal 30 has received the last chunk of the content k, the content reproduction recording unit 26 adds 1 to the number of completions of reproduction of the content k in the content management table 213. When the user terminal 30 stops the playback in the middle of the chunk number 2 of the content k, the content playback recording unit 26 adds 1 to the number of playback stops of the chunk number 2 of the content k in the cache chunk information table 212.

  According to the content distribution system described above, the cache server 20 holds chunks of content that are highly likely to be viewed by the user terminal 30. In addition, when each cache server 20 runs out of its own cache capacity (the storage capacity of the cache server storage unit 21), it deletes chunks that are less likely to be viewed from chunks of content held by itself. As a result, the cache server 20 holds chunks of content that are more likely to be viewed by the user terminal 30, so that the cache efficiency in the cache server 20 can be improved.

<Processing procedure>
Next, the processing procedure of the content distribution system will be described with reference to FIGS.

  First, the procedure in which the origin server 10 delivers content (content chunk) to the cache server 20 will be described with reference to FIG. This content distribution is preferably performed in a time zone in which the traffic volume between the origin server 10 and the cache server 20 is relatively small. The origin server 10 obtains the viewing leaving characteristics (see FIG. 2) for each content group from the viewing leaving characteristics and content attributes of each content measured in advance (S1 in FIG. 5). Then, the origin server 10 obtains a reproduction time T (Gx, α) corresponding to the viewing leaving rate α for the content group Gx (x = 0, 1,...) From the viewing leaving characteristics obtained in S1 (S2). It is registered in the viewing separation characteristic table 152 (see Table 1) (S3).

  The prior cache target selection unit 11 (see FIG. 3) of the origin server 10 refers to the viewing / leaving characteristic table 152 and determines to which chunk each content is distributed to the cache server 20 (S4). Specifically, the pre-cache target selection unit 11 refers to the viewing / leaving characteristic table 152 and acquires T (Gx, α) of the content of each content group. Then, the pre-cache target selection unit 11 refers to the chunk information table 153 of the content (for example, content k) of each content group, and the chunk whose playback time length from the first chunk exceeds T (Gx, α) Select up to the previous cache target chunk.

  Then, the content transmission unit 12 of the origin server 10 transmits an inquiry about the free capacity (the free capacity for storing the pre-cache target chunk (the chunk of the content to be distributed) selected in S4) to the cache server 20 (S5). ). Note that the content transmission unit 12 transmits the free capacity inquiry including the data size of the pre-cache target chunk.

  When the origin server communication unit 22 (see FIG. 4) of the cache server 20 receives this free capacity inquiry, it confirms with the cache management unit 23 whether there is free capacity in the cache server storage unit 21 (S6). . In other words, the origin server communication unit 22 outputs the data size of the pre-cache target chunk included in the free capacity inquiry received from the origin server 10 to the cache management unit 23. Based on the data size of the output pre-cache target chunk, the cache management unit 23 checks whether or not the cache server storage unit 21 has a free capacity for storing the pre-cache target chunk. When the cache management unit 23 determines that there is no free capacity in the cache storage unit 21 (No in S6), the cache management unit 23 creates the necessary free capacity among the content chunks 211 in the cache server storage unit 21. A chunk to be deleted is selected, and the selected chunk is deleted (S7). Then, the process proceeds to S8. Details of the process of S7 will be described later.

  On the other hand, when the cache management unit 23 determines that the cache server storage unit 21 has free space (Yes in S6), the origin server communication unit 22 of the cache server 20 transmits a cache OK response to the origin server 10. (S8).

  The content transmission unit 12 of the origin server 10 that has received the cache OK response from the cache server 20 sends the chunk (cache target chunk) determined in S4, the chunk information of this chunk, and the content information of the cache target chunk to the origin server. The data is read from the storage unit 15 and distributed to the cache server 20 (S9).

  The origin server communication unit 22 of the cache server 20 adds the stored chunk to the content chunk 211 of the cache server storage unit 21 via the cache management unit 23 and stores it. Further, the origin server communication unit 22 stores the chunk information distributed via the cache management unit 23 in the cache chunk information table 212 and stores the content information in the content management table 213 (S10).

  In this way, the cache server 20 holds chunks of content up to the playback time T (Gx, α) that is the viewing withdrawal rate α.

  Next, the processing procedure of the cache server 20 that has received the content distribution request from the user terminal 30 will be described with reference to FIG.

  The user terminal 30 transmits a content distribution request to the cache server 20 (S11). When the request reception unit 24 (see FIG. 4) of the cache server 20 receives the content distribution request, the content distribution unit 25 refers to the content management table 213 (see Table 5) and registers the requested content. It is confirmed whether or not (S12). If the content is registered (Yes in S13), the content distribution unit 25 starts distributing the chunk of content to the user terminal 30 that has transmitted the content distribution request (S14). Then, the process proceeds to S15.

  On the other hand, when there is no content registration in S13 (No in S13), the origin server communication unit 22 transmits a content request for the requested content to the origin server 10 (S31). And the content transmission part 12 of the origin server 10 which received the content request delivers the chunk of the requested content (S32). Then, the process proceeds to S14.

  After S14, the content reproduction / recording unit 26 of the cache server 20 adds 1 to the number of accesses to the content in the content management table 213 (see Table 5) (S15).

  Then, the content distribution unit 25 checks whether or not the number of undelivered chunks in the content chunk 211 of the content in the cache server storage unit 21 is equal to or less than a threshold value (subsequent chunk request threshold value) (S16). That is, it is confirmed whether or not there are only a few chunks of content held in the cache server storage unit 21. Here, when the number of undelivered chunks is equal to or less than a predetermined threshold (subsequent chunk request threshold) (Yes in S16), the content distribution unit 25 transmits the subsequent chunks to the origin server 10 via the origin server communication unit 22. (Succeeding chunk of the chunk held by itself) is requested (S17). Then, the process proceeds to S18 and S21. On the other hand, when the number of undelivered chunks is not less than the predetermined threshold (subsequent chunk request threshold) (No in S16), the process proceeds to S21.

  Upon receiving the chunk request (request) from the cache server 20, the content transmission unit 12 of the origin server 10 delivers the requested chunk to the cache server 20 (S18), and the origin server communication unit 22 of the cache server 20 delivers the chunk. The cached chunk is cached (S19). That is, the origin server communication unit 22 stores the distributed chunks in the cache server storage unit 21 via the cache management unit 23. Then, the origin server communication unit 22 registers the storage location (cache storage location) of this chunk in the cache chunk information table 212 (see Table 4) (S20). At this time, the origin server communication unit 22 also registers the chunk information of the delivered chunk in the cache chunk information table 212 (see Table 4). Then, the process ends.

  In S21, the content distribution unit 25 of the cache server 20 receives the chunk request from the user terminal 30 (Yes in S21), and distributes the chunk to the user terminal 30 (S22). Then, the process proceeds to S23. On the other hand, if there is no chunk request from the user terminal 30 (No in S21), the content distribution unit 25 returns to S21. In S23, when the cache management unit 23 of the cache server 20 confirms that the content has been delivered to the last chunk (Yes in S23), the content management table 213 (see Table 5) adds 1 to the number of completed playbacks of the content (see Table 5) (S23). S24). On the other hand, when the content has not yet been distributed to the last chunk (No in S23), the process returns to S16.

  In this way, the cache server 20 distributes the chunk of content to the user terminal 30.

  Next, a processing procedure when the user terminal 30 stops the reproduction of content will be described with reference to FIG. When the user terminal 30 stops the reproduction of the content (S41), the user terminal 30 notifies the cache server 20 that is the content distribution source of the reproduction stop request or the session end.

  When the content distribution unit 25 (see FIG. 4) of the cache server 20 receives the reproduction stop request or the session end from the user terminal 30 or detects the session disconnection, the content distribution unit 25 stops the content distribution (S42). Then, the content reproduction recording unit 26 acquires the chunk number of the chunk of the content that has been reproduced (distributed) when the reproduction is stopped, the session is terminated, or the session is terminated, and the corresponding chunk number in the cache chunk information table 212 (see Table 4). 1 is added to the number of reproduction stops (S43).

  In this way, in the cache chunk information table 212 of the cache server 20, it is recorded which chunk of each content chunk has been reproduced (received) by the user terminal 30. The information in the cache chunk information table 212 is referred to when the cache management unit 23 selects a chunk of the content chunk 211 to be preferentially deleted when the free capacity of the cache server storage unit 21 is reduced.

  Here, when the cache server 20 does not have the free capacity S necessary for storing the chunk from the origin server 10, the cache management unit 23 performs the following procedure to store the corresponding chunk until the free capacity S is secured. Delete from the content chunk 211. The deletion procedure at this time will be described using FIG. 8 with reference to FIG. 4 as appropriate. Note that the initial values of the parameters N1 and N2 used here are predetermined values N1i and N2i (N1i ≧ N2i), respectively.

  The cache management unit 23 refers to the content management table 213 (see Table 5) and the cache chunk information table 212 (see Table 4) and is delivering content to the user terminal 30 (between receiving the content delivery request and ending delivery). Instead, the content chunk 211 of the content to be preferentially deleted is selected for content that has passed a predetermined time or more from the content reception date and time. First, referring to the content management table 213 and the cache chunk information table 212, the total chunk data amount S1 of the content with the access count = 0 among the content chunks 211 of the corresponding content is calculated (S51). If S> S1 (Yes in S52), the corresponding data is deleted, (S-S1) is set to the value of S (S53), and the process proceeds to S54. That is, if there is content that has never been viewed by the user terminal 30, the content is deleted. Further, for example, in the content management table 213 (see Table 5), a content distribution flag (illustrated) is shown in the content management table 213 (see Table 5) during the period from the reception of the content distribution request to the end of distribution so that the cache server 20 can determine whether the content is being distributed. (Omitted) is turned on. Thereby, the cache management unit 23 can prevent the content being distributed from being accidentally deleted. Hereinafter, when the cache management unit 23 deletes the data of the content chunk 211, the content is transferred to the user terminal 30 with reference to the content management table 213 (see Table 5) and the cache chunk information table 212 (see Table 4). Content that is being distributed (between receiving the content distribution request and ending distribution) is excluded from deletion.

  On the other hand, if S> S1 is not satisfied (No in S52), the content with the number of accesses = 0 is deleted (S70). That is, the cache management unit 23 deletes the chunk data of the content with the access count = 0 in order from the content with the oldest content reception date and time, and sets the value obtained by subtracting the deleted chunk data amount from S to the value of S. Until the S becomes 0 or less, the content deletion is repeated in ascending order of the content reception date and the process is terminated.

  In S <b> 54, the cache management unit 23 refers to the content management table 213 and the cache chunk information table 212, and if there is content that satisfies the content access number N <b> 1 or less and the reproduction completion number = 0 (Yes in S <b> 54), S <b> 55. Perform the process. On the other hand, if there is no content satisfying N1 or less and the number of completed reproductions = 0 (No in S54), the process proceeds to S58. S58 will be described later.

  In S55, the cache management unit 23 refers to the content management table 213 and the cache chunk information table 212, and the withdrawal rate is β (predetermined value) for each content that satisfies the content access number N1 or less and the playback completion number = 0. Chunk number c exceeding) is calculated. Here, the withdrawal rate is a value defined by the sum of the number of reproduction stops from the reproduction start of the content to the chunk / the number of accesses to the content (number of distributions). Then, the cache management unit 23 calculates the total data amount S2 of the chunks after the chunk number c + 1 for all the target contents, deletes the corresponding data (that is, the chunks after the chunk number c + 1 of the target contents), ( S-S2) is set to the value of S (S56).

  After S56, if S> 0 (Yes in S57), the process proceeds to S58. On the other hand, if S> 0 is not satisfied (No in S57), the process is terminated.

  In S58, the cache management unit 23 refers to the content management table 213, and if there is content that satisfies the content access number N1 or less and the reproduction completion number N2 or less (Yes in S58), the cache management unit 23 Referring to the content management table 213 and the cache chunk information table 212, the chunk number c whose withdrawal rate exceeds β (predetermined value) is calculated for each content (S59). On the other hand, in S58, if there is no content that satisfies the content access number N1 or less and the reproduction completion number N2 or less (No in S58), the process proceeds to S62.

  After S59, the cache management unit 23 refers to the cache chunk information table 212, calculates the total data amount S3 of the chunks after the chunk number c + 1 for all target content, and applies the corresponding data (that is, the target content (Chunks after chunk number c + 1) are deleted, and (S-S3) is set to the value of S (S60).

  After S60, if S> 0 (Yes in S61), the process proceeds to S62. On the other hand, if S> 0 is not satisfied (No in S61), the process is terminated.

  In S62, when the value obtained by adding the predetermined value A1 to N1 is not less than N1max (the maximum value set in N1) (No in S62), the cache management unit 23 adds the value of N2 to the predetermined value A2. (S63), if N2 ≦ N1 (Yes in S64), the process proceeds to S58. . When N2 ≦ N1 is not satisfied (No in S64), the cache management unit 23 adds the value of N1max to the predetermined value A3, sets the value of N2 to N2i (S65), and returns to S62.

  On the other hand, when N1 + A1 is equal to or smaller than N1max (the maximum value set in N1) in S62 (Yes in S62), the cache management unit 23 adds A1 to the value of N1 (S66), and returns to S54.

  As described above, when the cache server 20 reserves a free space in the cache server storage unit 21, the cache server 20 preferentially selects, from the content chunk 211, the content with the smaller number of distributions and the number of completed reproductions as the data to be deleted. Then, the cache server 20 determines that the withdrawal rate (sum of the number of reproduction stops from the start of reproduction of the content to the chunk / the number of accesses to the content (number of distributions)) of the selected content chunks is a predetermined threshold value β. The data of the chunks after the time exceeding is deleted until the free capacity for caching the content data from the origin server 10 can be secured. As a result, chunks that are likely to be viewed (reproduced) by the user terminal 30 remain in the cache server storage unit 21, so that the cache efficiency of the cache server 20 can be improved.

  The selection of the data to be deleted may be performed at the timing when the cache management unit 23 of the cache server 20 receives a new content chunk delivery notification from the origin server 10, or the chunk delivery notification. Regardless of, you may go in advance.

  For example, the cache management unit 23 performs, for each predetermined period, the number of accesses and the number of playback completions of each content shown in the content management table 213 (see Table 5) and each of the numbers shown in the cache chunk information table 212 (see Table 4). A list (cache deletion list) indicating in advance which chunks should be deleted preferentially when the free capacity in the cache server storage unit 21 is insufficient is created by referring to the number of content chunks to be stopped. deep.

  Here, the cache management unit 23 selects, as data to be deleted preferentially, (1) a content with the smallest number of accesses and (2) a content with the smallest number of playback completions. Of the chunks constituting the selected content, (3) a chunk whose withdrawal rate (the sum of the number of playback stops from the start of playback of the content to the chunk / the number of accesses to the content (number of distributions)) is β Select the chunk with the previous chunk number as the chunk to be deleted. That is, in selecting a chunk to be deleted preferentially by the cache management unit 23, (1) a small number of accesses, (2) a small number of playback completions, and (3) a high withdrawal rate. Three conditions are used. Here, the priority order of condition application is, for example, (1) → (2) → (3).

  When the cache management unit 23 creates the cache deletion list in this way, the cache management unit 23 stores it in a predetermined area of the cache server storage unit 21. When the origin server communication unit 22 of the cache server 20 receives a new chunk of content from the origin server 10 and the cache server storage unit 21 has insufficient free space, it is indicated in this cache deletion list. In accordance with the priority order, the chunk is deleted from the cache server storage unit 21 until a free space for a new content chunk can be secured.

  By doing so, the cache server 20 immediately deletes the free space even if the cache server storage unit 21 has insufficient free space to store the new content chunk, Can be prepared for reception.

  In the embodiment described above, the subsequent chunk request threshold used when the cache server 20 requests a subsequent chunk to the origin server 10 may use a common value for each content, or set an individual value for each content. You may make it use. When using individual values, the subsequent chunk request threshold for each content is set in the content management table 213.

  The origin server 10 may further include a viewing / leaving characteristic table creation unit 13 indicated by a broken line. The viewing leaving characteristic table creating unit 13 calculates the viewing leaving characteristic for each content group based on the distribution result of the content 151 stored in the origin server storage unit 15 to each cache server 20, and the calculated viewing leaving characteristic. The viewing leaving characteristic table 152 is created based on the above. Then, the viewing / leaving characteristic table creating unit 13 updates the viewing / leaving characteristic table 152 based on the distribution result of the content to each cache server 20 for each predetermined period. For example, when the origin server 10 delivers a chunk of content in response to a request from the cache server 20, it is recorded in the content information table 154 which chunk has been delivered up to which chunk. Then, the viewing leaving characteristic table creating unit 153 creates or updates the viewing leaving characteristic table 152 based on the distribution result of the chunk of the content. In this way, the content distribution result from the origin server 10 to the cache server 20 can be reflected in the viewing withdrawal characteristic table 152.

DESCRIPTION OF SYMBOLS 10 Origin server 11 Prior cache object selection part 12 Content transmission part 13 Viewing leaving characteristic table preparation part 15 Origin server memory | storage part 20 Cache server 21 Cache server memory | storage part 22 Origin server communication part 23 Cache management part 24 Request receiving part 25 Content delivery part 26 Content reproduction recording unit 30 User terminal 151 Content 152 Viewing / leaving characteristic table 153 Chunk information table 154 Content information table 211 Content chunk 212 Cache chunk information table 213 Content management table

Claims (5)

A content delivery system comprising a cache server that delivers video content in response to a request from a user terminal, and an origin server that delivers the content to the cache server,
The origin server is
The content information table indicating the content group to which the content belongs for each content, and the viewing leaving rate defined by the following equation (1) for each content group Gx of the content is a predetermined value α. An origin server storage unit for storing a viewing / leaving characteristic table indicating a reproduction time T (Gx, α) of content of the content group Gx,
Viewing disconnection rate at time t of playback of content = (number of views at the start of playback of the content−number of views at the time of playback of the content at t seconds) / (number of views at the start of playback of the content) 1)
Referring to the viewing / leaving characteristic table and the content information table, content that distributes data from the beginning until the playback time T (Gx, α) is exceeded to the cache server among the content data of the content group Gx A transmission unit,
The cache server is
A cache server storage unit for storing data of the distributed content;
A content distribution unit that distributes the data of the content to the user terminal that has transmitted the request in response to a request from the user terminal;
A content distribution system comprising: an origin server communication unit configured to transmit a distribution request for the content data to the origin server when the content data requested from the user terminal is not in the cache server storage unit. . The cache server storage unit further includes:
Store the content management table,
The cache server further includes:
A cache for determining whether or not the cache server storage unit has a free capacity for caching the content data from the origin server based on the data size of the content data to be distributed included in the inquiry from the origin server The management department,
A content reproduction recording unit that records the number of distributions of the content to the user terminal and the number of reproduction completions in the user terminal for each content in the content management table;
The origin server communication unit further includes:
When an inquiry is received from the origin server as to whether or not there is a free capacity for caching the content data in the cache server storage unit, the data size of the content data to be distributed included in the inquiry is set to the cache Output to the management unit, and when the cache management unit determines that there is a free capacity to cache the content data, sends a delivery request for the content data to the origin server;
The cache management unit
When it is determined that there is no free capacity to cache the content data from the origin server, the content management table is referred to as the data to be deleted, and the content is already stored in the cache server storage unit. 2. The content distribution according to claim 1, wherein the process of deleting the selected content is repeated until priority is selected from the content with a small number of distributions and the number of completed reproductions until the free space can be secured. system. The content playback / recording unit further includes:
When there is a stop of playback of the content, record up to what data of the content has been recorded in the cache chunk information table,
The cache server storage unit further includes:
Storing the cache chunk information table;
The cache management unit
When it is determined that there is no free capacity to cache the content data from the origin server, the content management table and the cache chunk information table are referred to as the data to be deleted and are already stored in the cache server storage unit Searching for content with the number of distributions of 0 from the content, and deleting the content with the number of distributions of 0, but when the free space cannot be secured, priority is given to content with a smaller number of distributions and playback completions of the content To obtain data of a portion after the time when the departure rate from the reproduction start defined by the following formula (2) exceeds a predetermined threshold β among the data of the selected content The process of deleting the obtained data is repeated until the data can be secured. Content delivery system according to claim 2.
Exit rate of the content at the playback time t seconds = Number of playback cancellations from the start of playback of the content to the playback time t seconds / Number of times the content is distributed. Formula (2)   The content distribution system according to any one of claims 1 to 3, wherein the content transmission unit and the content distribution unit distribute the data of the content in a chunk format. A content distribution method in a content distribution system comprising: a cache server that distributes video content in response to a request from a user terminal; and an origin server that distributes the content to the cache server,
The content information table indicating the content group to which the content belongs for each content, and the viewing leaving rate defined by the following equation (1) for each content group Gx of the content is a predetermined value α. The origin server comprising an origin server storage unit that stores a viewing / leaving characteristic table indicating the playback time T (Gx, α) of the content of the content group Gx,
Viewing disconnection rate at time t of playback of content = (number of views at the start of playback of the content−number of views at the time of playback of the content at t seconds) / (number of views at the start of playback of the content) 1)
Distributing to the cache server data from the beginning of the content data of the content group Gx until the playback time T (Gx, α) is exceeded with reference to the viewing / leaving characteristic table and the content information table Run
The cache server is
Storing the distributed content data in a cache server storage unit;
Delivering the content data to the requesting user terminal in response to a request from the user terminal;
When the content data requested from the user terminal is not in the cache server storage unit, a step of transmitting a content data distribution request to the origin server is executed.

Images