JP2011239069A - Content distribution method, peer node and program which use overlay network and adapt to user preference - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method of distributing content in real time only to users that are interested in the content, without installing a particular content distribution server on a network.SOLUTION: A plurality of peer nodes are connected to an overlay network virtually configured. Content to transmit has a feature amount vector based on its substance. Each peer node has a routing table in which preference vectors of users that possess a neighboring peer node capable of achieving communication by one hop are registered. The peer nodes includes a preference similarity determination means which compares the feature amount vector of the content and the preference vectors of each neighboring peer node registered in the routing table and determines whether similarity is within the prescribed similarity scope or not, and a content transmission means which transmits the content to the neighboring peer nodes based on the preference vectors when the similarity is determined to be within the prescribed similarity scope.

Description

  The present invention relates to a technology for distributing content posted or updated in real time to a user terminal in accordance with user preferences.

  Recently, the content distributed via the Internet is not limited to that created by a company, and there is also CGM (Consumer Generated Media) content such as blogs, videos, and images created by users. Such CGM content is uploaded from the terminal to the CGM site by a user operation. The other user can use the terminal to search the CGM site with a keyword and browse / view desired CGM content. Typical CGM sites include Amoeba Blog (registered trademark) (http://ameblo.jp/), Nico Nico Douga (registered trademark) (http://www.nicovideo.jp/) or Youtube (registered trademark) (http http://www.youtube.com/?gl=JP&hl=en).

  Further, there is an RSS (RDF Site Summary) reader as a technique for distributing CGM content posted by a user only to users who are interested in the content in real time. According to this technique, the user registers the CGM site in advance with the terminal, so that the terminal periodically confirms the update of the site. Thereby, the user can acquire the latest information in almost real time. However, even if the content does not match the content desired by the user, the content is distributed if the site is registered in advance. On the contrary, even if the content is a content desired by the user, the content is not distributed unless it is registered in advance.

  On the other hand, a blog distribution system that calculates the similarity between the content of a new blog article (content) and user-desired information registered in the distribution server, and distributes the content preferentially to users with a high degree of similarity (For example, refer nonpatent literature 1). According to this technology, it is possible to distribute only the content of the contents desired by the user in real time without registering the site in advance.

According to this technique, for example, the following sequence is adopted.
(S1) The blog distribution server periodically acquires new blog articles from commercial blog sites.
(S2) The blog distribution server calculates a feature vector in the blog article using the category dictionary.
(S3) The blog distribution server calculates a Euclidean distance between a user preference vector registered in advance and a feature amount vector of the content, and distributes the content preferentially to a user with a small Euclidean distance.

Hajime Hattori, KDDI R & D Labs, “Destination User Fast Selection Method for Real-Time Blog Distribution”, IEICE, Web Intelligence and Interaction Study Group Report, [online], [April 11, 2010 Search], Internet <URL: http: //www.ieice.org/~wi2/pastcfa/pastcfa14.html> "A Scalable Content Addressable Network", Proceedingof ACM SIGCOMM 2001

  According to the technique described in Non-Patent Document 1, the blog distribution server centrally manages content, so that the scale increases as the number of contents and the number of users increase. In order to calculate a feature vector for each content (new blog article) and calculate a Euclidean distance between the feature vector and the preference vector, the processing amount also increases in proportion to the number of contents and the number of users. In order to require such processing, according to this technology, a service that delivers information on the content to a user's terminal that wants the content immediately after the content is posted is not assumed.

  Accordingly, an object of the present invention is to provide a content distribution method, a peer node, and a program capable of distributing content in real time only to users who are interested in the content without installing a specific content distribution server in the network. And

According to the present invention, there is provided a content distribution method in an overlay network virtually configured on a physical network connected by a plurality of peer nodes,
The content to be transmitted is assigned a feature vector based on the content,
Each peer node has a routing table that registers an identification value of an adjacent peer node that can communicate in one hop, and a preference vector of a user possessing the adjacent peer node,
A determination step in which the peer node compares the feature vector of the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table, and determines whether the similarity is within a predetermined similarity range; ,
A transmission step of transmitting content to an adjacent peer node based on a preference vector when it is determined that the similarity is within a predetermined similarity range.

According to another embodiment of the content distribution method of the present invention,
An overlay network is configured by arranging peer nodes in a multidimensional space (n dimension: n ≧ 2),
It is also preferable that the routing table registers the previous and next peer nodes as adjacent peer nodes for each dimension i (i = 1 to n).

  According to another embodiment of the content distribution method of the present invention, the overlay network is preferably a CAN (Content Addressable Network).

According to another embodiment of the content distribution method of the present invention, for the previous stage of the determination step,
A receiving step in which the peer node receives content from an adjacent peer node in the i-dimensional direction;
A first selection step of selecting an adjacent peer node in dimensions 1 to (i-1) and (i + 1) as a transmission destination of the content;
For the i dimension, it is also preferable to include a second selection step of selecting an adjacent peer node opposite to the content transmission source adjacent peer node as a transmission destination of the content.

According to another embodiment of the content distribution method of the present invention,
A sequence number is assigned to the data packet of the content to be transferred,
The first selection step is
The peer node transmits a transmission confirmation request including a sequence number to an adjacent peer node selected as a transmission destination of the content;
When the adjacent peer node receives the transmission confirmation request, it compares the sequence number of the content already received with the sequence number included in the transmission confirmation request, and returns a transmission confirmation response including the comparison result to the peer node; ,
The peer node preferably includes a step of determining whether or not to transmit to the adjacent peer node based on the comparison result when the transmission confirmation response is received.

  According to another embodiment of the content distribution method of the present invention, each peer node obtains a preference vector of the neighboring peer node by periodically sending a Keep-Alive message to the neighboring peer node registered in the routing table. It is also preferable to do.

According to another embodiment of the content distribution method of the present invention, for the previous stage of the determination step,
A content distribution source node extracting from the content to be transmitted a feature vector based on the content;
A content distribution source node deriving a destination identification value based on the feature quantity vector;
The content distribution source node sending a search request to the peer node of the destination identification value;
It is also preferable that the content distribution source node has a step of transmitting the content toward the destination address included in the search response when the search response is received.

According to the present invention, a peer node connected to a virtually configured overlay network on a physical network connected by a plurality of peer nodes,
The content to be transmitted is assigned a feature vector based on the content,
Each peer node has an identification value of an adjacent peer node that can communicate in one hop, and a routing table in which a preference vector of a user possessing the adjacent peer node is registered,
A preference similarity determination unit that compares the feature vector of the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table and determines whether the similarity is within a predetermined similarity range; ,
When it is determined that the similarity is within a predetermined similarity range, content transmitting means for transmitting content to an adjacent peer node based on a preference vector is provided.

According to another embodiment in the peer node of the present invention,
The overlay network is configured by arranging peer nodes in a multidimensional space (n dimension: n ≧ 2),
It is also preferable that the routing table registers the previous and next peer nodes as adjacent peer nodes for each dimension i (i = 1 to n).

  According to another embodiment in the peer node of the present invention, the overlay network is also preferably a CAN.

According to another embodiment in the peer node of the present invention,
content receiving means for receiving content from adjacent peer nodes in the i-dimensional direction;
First transmission destination selection means for selecting adjacent peer nodes in the 1 to (i-1) and (i + 1) dimensions as transmission destinations of the content;
For the i dimension, it is preferable to further include a second transmission destination selection means for selecting an adjacent peer node opposite to the content transmission source adjacent peer node as the transmission destination of the content.

According to another embodiment in the peer node of the present invention,
A sequence number is assigned to the data packet of the content to be transferred,
The first transmission destination selection means is
A transmission confirmation request transmission function for transmitting a transmission confirmation request including a sequence number to an adjacent peer node selected as a transmission destination of the content;
A transmission confirmation response reply function that compares the sequence number of the content already received with the sequence number included in the transmission confirmation request when a transmission confirmation request is received, and returns a transmission confirmation response including the comparison result to the peer node; ,
It is also preferable to have a transmission determination function for determining whether to transmit to the adjacent peer node based on the comparison result when the transmission confirmation response is received.

  According to another embodiment of the peer node of the present invention, the neighboring peer confirmation means for acquiring the preference vector of the neighboring peer node by periodically sending a Keep-Alive message to the neighboring peer node registered in the routing table. It is also preferable to have it.

According to another embodiment in the peer node of the present invention,
Having a content distribution means as a content distribution source;
Content distribution means
A feature vector extraction function for extracting a feature vector based on the content from the content to be transmitted;
A destination identification value deriving function for deriving a destination identification value based on the feature vector;
A search request transmission function for transmitting a search request to the peer node of the destination identification value;
A search response receiving function for receiving a search response;
It is also preferable to have a content distribution function for distributing content toward the destination address included in the search response.

According to the present invention, a program for causing a computer mounted on a peer node connected to a virtually configured overlay network to function on a physical network connected by a plurality of peer nodes,
The content to be transmitted is assigned a feature vector based on the content,
Each peer node has an identification value of an adjacent peer node that can communicate in one hop, and a routing table in which a preference vector of a user possessing the adjacent peer node is registered,
A preference similarity determination unit that compares the feature vector of the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table and determines whether the similarity is within a predetermined similarity range; ,
When it is determined that the similarity is within a predetermined similarity range, the computer is caused to function as a content transmission unit that transmits content to an adjacent peer node based on a preference vector.

  According to the content distribution method, peer node, and program of the present invention, the content is distributed while using the overlay network and determining the similarity between the content feature vector and the user preference vector. As a result, content can be distributed in real time only to users who are interested in the content without installing a specific content distribution server on the network.

1 is a system configuration diagram of an overlay network. It is explanatory drawing showing the logical structure of the overlay network which has arrange | positioned the peer node in multidimensional space. It is explanatory drawing of the two-dimensional space which transfers a content to the destination node based on the feature-value vector. It is a flowchart of this invention in FIG. It is a flowchart which delivers a content to the adjacent peer node in an overlay network. It is explanatory drawing of the two-dimensional space showing the transfer of FIG. It is a flowchart which selects the adjacent peer node used as the transmission destination of a content. It is a functional block diagram of the peer node in this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  According to the present invention, an overlay network technology in which user terminals (peer nodes) are virtually interconnected is used for content distribution. An overlay network is a network in which peer nodes are virtually interconnected by covering (overlaying) an IP (Internet Protocol) layer in a physical network. As a typical overlay network, for example, there is a P2P (Peer to Peer) network.

  FIG. 1 is a system configuration diagram of an overlay network.

  According to the system of FIG. 1, a plurality of peer nodes 1 are connected to a physical network 2. On the physical network 2, an overlay network 3 is virtually configured by a plurality of peer nodes 1. Here, the peer node is a terminal possessed by the user and may be a personal computer or a home terminal such as a set top box or a home gateway. For example, the source peer node and the destination peer node may function as SIP-UA (SIP-User Agent), and the overlay network may be constructed by P2PSIP (Peer to Peer-Session Initiation Protocol).

  As a technique of structured overlay, there is “CAN (Content Addressable Network)” using a multidimensional orthogonal coordinate space (for example, see Non-Patent Document 2). According to this technology, unlike the conventional P2P technology represented by Napster or Gunutella, it is possible to search data distributed and arranged in the entire network at high speed. In addition, the network can be reconstructed in an autonomous and distributed manner with respect to frequent participation and withdrawal of peer nodes.

  FIG. 2 is an explanatory diagram showing a logical configuration of an overlay network in which peer nodes are arranged in a multidimensional space.

  According to the overlay network of the present invention, peer nodes are arranged in a multidimensional space (n dimensions: n ≧ 2). The entire coordinate space is divided into a plurality of zones, and each peer node manages one divided zone. Each peer node maintains a “routing table” and manages information of user terminals (adjacent peer nodes) existing in a zone adjacent to its own zone.

  The “routing table” registers peer nodes before and after the adjacency for each dimension i (i = 1 to n). The routing table in the present invention registers at least a “preference vector” of a user having an adjacent peer node that can communicate with one hop.

Table 1 below shows the configuration of the routing table held by each peer node for the two-dimensional case.

“Node number”: A unique number on the overlay network “Adjacent dimension”: The dimension that the adjacent peer node is adjacent to when viewed from the own node “Adjacent direction”: The adjacent peer node in the adjacent dimension is adjacent when viewed from the own node Direction (Forward (F: Forward) / Reverse (B: Backward))
"IP address"
"port number"
“Identification value”: Position represented by a plurality of dimensions i (1 to n) “Zone information”: Zone range managed by the adjacent peer node “Preference vector”: Vector notation indicating the preference of the user possessing the adjacent peer node

  Each peer node preferably acquires information on the adjacent peer node by periodically transmitting a Keep-Alive message to the adjacent peer node registered in the routing table.

  According to FIG. 2, a two-dimensional plane space is represented as a multidimensional space. Each planar space is divided into a plurality of zones. According to FIG. 2, the preference value of each genre is digitized in the range of 0 to 1, and the value of the preference value is used as it is for the identification value on the overlay network.

According to FIG. 2, the peer node (user terminal) 17 is expressed as follows.
Preference value for genre A (one-dimensional identification value): 0.50
Preference value for genre B (two-dimensional identification value): 0.52

Each peer node generates an identification value of its own node based on the user preference vector when participating in the overlay network. The identification value includes an identification value in the one-dimensional direction and an identification value in the two-dimensional direction. Thereby, one zone corresponding to the identification value is assigned. According to FIG. 2, the following rectangular zones are assigned to the peer node 17.
Identification value of own node: [0.50, 0.52]
Zone information: [0.4 to 0.6, 0.4 to 0.6]
(1D range is 0.4 to 0.6, 2D range is 0.4 to 0.6)

Further, according to FIG. 2, for example, the peer node 17 has the following routing table.

  The “preference vector” is obtained by quantifying user preferences for each category [element] (for example, entertainment, society, economy,...). Note that the preference vector may be one in which the element value of the preference vector is set freely by the user, or the average value of the feature vector of the read content may be set.

  On the other hand, for the content to be distributed, a feature vector is calculated by content analysis based on the category dictionary. The feature vector also has the same elements as the preference vector. For example, when the content is text such as a blog article, text corresponding to each category is registered in the category dictionary. After extracting a large number of words from the content, the appearance frequency of words of all categories is calculated, and a feature vector is calculated.

FIG. 3 is a flowchart for transferring to the destination node based on the feature quantity vector of the content.
FIG. 4 is an explanatory diagram of a two-dimensional space representing the transfer of FIG.

(S31) For the content distribution source node, the content to be distributed is selected according to the user operation. Here, it is assumed that the content distribution source node is the peer node ID8.
(S32) The content distribution source node extracts a feature vector based on the content from the content to be transmitted by the content distribution source node.
(S33) Next, the content distribution source node derives a destination identification value in the overlay network based on the feature quantity vector. For example, according to FIG. 3, it is assumed that the destination identification value is included in the zone managed by the peer node ID 17.
(S34) The content distribution source node determines the sequence number of the content to be distributed so as to be unique within the network. This sequence number is for preventing distribution of duplicate contents. For example, it can be generated from information such as the IP address, port number, and time in the content distribution source node.
(S35) The content distribution source node transmits a search request (Search Request) including the destination identification value to the overlay network. The search request is transferred to the peer node having the vertex closest to the destination identification value. According to FIGS. 3 and 4, the search request transmitted from the peer node ID8 is repeatedly transferred through the peer nodes of ID9, ID10, and ID11. As a result, the user approaches the peer node that manages the destination identification value. Finally, the search request reaches the peer node ID 17 that manages the destination identification value.
(S36) The peer node ID 17 returns a search response (Search Response) to the content distribution source node. The search response includes the IP address and port number of the peer node ID 17.
(S37) The content distribution source node (peer node ID 8) transmits a transmission confirmation request (Confirm Request) to the peer node ID 17 using the IP address and port number included in the search response. The transmission confirmation request includes the sequence number determined in S34.
(S38) The peer node ID 17 that has received the transmission confirmation request executes reception determination. Specifically, it is determined whether or not the sequence number specified in the transmission confirmation request matches the sequence number in the list of already received contents. If they match, it is determined as NG, and if they do not match, it is determined as OK. A transmission confirmation response including the determination result is returned to the content distribution source node.
(S39) The content distribution source node (peer node ID 8) transmits the content to the peer node ID 17 when the determination result is OK. The distributed content includes a feature vector. Here, a similarity range representing a distribution range may be included. The similarity range will be described later with reference to FIG. If the determination result is NG, the content is not transmitted.

FIG. 5 is a flowchart for distributing content to adjacent peer nodes in the overlay network.
FIG. 6 is an explanatory diagram of a two-dimensional space representing the transfer of FIG.

  Through S39 of FIG. 3 described above, the peer node ID 17 has received the content from the content distribution source node (peer node ID 8). The peer node ID 17 accumulates content and updates the received content list. The received content list includes at least a file name, a distribution source address, a sequence number, a feature vector, a similarity range, and a reception time.

(S51) The peer node ID 17 selects an adjacent peer node to which content is to be distributed, using the routing table. Here, the peer node compares the feature vector of the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table, and determines whether the similarity is within a predetermined similarity range. . Then, when it is determined that the similarity is within the predetermined similarity range, an adjacent peer node based on the preference vector is selected.

  FIG. 7 is a flowchart for selecting an adjacent peer node as a content transmission destination.

When the peer node receives content from an adjacent peer node in the i-dimensional direction, the following processing is executed.
(S71) When the content is received, it is determined whether or not the transmission source node is a content distribution source node. This can be determined by comparing the distribution source address added to the content with the transmission source address of the data packet that received the content.
(S72) When the distribution source address matches the transmission source address, all adjacent peer nodes registered in the routing table are selected as transmission destinations. According to FIGS. 5 and 6, since the peer node ID 17 has received the content from the peer node ID 8 which is the content distribution source node, all the adjacent peer node IDs 11, 16, 18 and 23 are selected.
(S73) When the distribution source address and the transmission source address do not match, the positional relationship of the transmission source node is determined by referring to the routing table. Specifically, the adjacent dimension (i) and the adjacent direction of the transmission source node are determined. The adjacent direction is determined to be forward when the identification value in the i-dimension of the user terminal of the content transmission source is larger than the identification value in the i-dimension of the own peer node, and is determined to be Backward when it is smaller.

(S74) Based on the adjacent dimension i of the transmission source node, all the adjacent peer nodes in the forward and reverse directions in the 1 to (i-1) and (i + 1) dimensions are selected as transmission destinations of the content.

  Here, in an environment where there is no participation or withdrawal of nodes, content can be sufficiently distributed by transmitting only to “1 to (i−1) dimensions”. For example, in the case of FIG. 6, since the node 16 has received the content from the one-dimensional direction, the content may be transmitted only to the terminal 15 that is the terminal opposite to the one-dimensional direction. However, in order to cope with an environment in which a node suddenly leaves the network, according to the present invention, content transmission is also confirmed in “(i + 1) dimension”. Therefore, although it can be said that it is redundant to transmit content also to the (i + 1) -dimensional node, this ensures reliable content distribution. On the other hand, the present invention does not redundantly distribute contents as compared with the case where “1 to (i−1) dimension and (i + 1) to n dimension” adjacent peer nodes are used as transmission destinations.

(S75) Next, for the adjacent dimension i, an adjacent peer node opposite to the transmission source node is selected as the transmission destination of the content. Here, if the source node is in Forward, only the Backward adjacent peer node is selected. On the other hand, when the transmission source node is in the Backward, only the Forward adjacent peer node is selected.
(S76) The processing of S77 to S79 is executed for each selected adjacent peer node.
(S77) The similarity between the content feature vector and the preference vector of each adjacent peer node registered in the routing table is calculated. The similarity may be a Euclidean distance between the feature vector and the preference vector.
(S78) Next, it is determined whether or not the similarity is within a predetermined similarity range. An adjacent peer node whose similarity is equal to or greater than the predetermined similarity range is not selected as a content transmission destination.
(S79) On the other hand, an adjacent peer node whose similarity is determined to be smaller than the predetermined similarity range is selected as a content transmission destination. Then, a transmission confirmation request is transmitted to the selected adjacent peer node.

  Returning again, reference is made to FIGS.

(S52) Since the transmission source node of the content is the content distribution source node, the peer node ID 17 selects all the adjacent peer nodes registered in the routing table as the transmission destination. According to FIG. 6, the peer node ID 17 selects ID11, ID16, ID18, and ID23 as the adjacent peer nodes of the content transmission destination. Then, the peer node ID 17 transmits a transmission confirmation request to these four adjacent peer nodes. In the transmission confirmation request, a sequence number is given for each content to be transferred.
(S53) On the other hand, the adjacent peer nodes of ID11, ID16, ID18, and ID23 return a transmission confirmation response to the peer node ID17. When the sequence number included in the transmission confirmation request matches the sequence number in the already received content, a transmission confirmation response including the determination result NG is returned.
(S54) The peer node ID 17 transmits the content to the adjacent peer nodes of ID 11, ID 16, ID 18, and ID 23 that have received the transmission confirmation response including the determination result OK.
(S55) Furthermore, the peer node with ID11 receives content from the peer node ID17 located in the two-dimensional Backward. At this time, the peer node of ID11 selects the one-dimensional Forward and Backward adjacent peer nodes (ID10 and ID12) and the two-dimensional Forward neighbor peer node (ID4) as content transmission destinations. Then, the peer node of ID11 transmits a transmission confirmation request to the adjacent peer nodes of ID10, ID12, and ID4.
(S56) On the other hand, the adjacent peer nodes of ID10, ID12, and ID4 return a transmission confirmation response to the peer node of ID11.
(S57) The peer node of ID11 transmits the content to the adjacent peer nodes of ID10, ID12, and ID4 that have received the transmission confirmation response including the determination result OK.

  The peer nodes ID16, ID18, and ID23 operate in the same manner as the peer node ID17.

(S58) Further, the peer node with ID 10 receives the content from the peer node ID 11 located in the one-dimensional Forward. At this time, the peer node of ID 10 selects the one-dimensional backward adjacent peer node ID 9 and the two-dimensional forward and backward adjacent peer node IDs 2 and 16 as the content transmission destinations. Here, it is assumed that the similarity between the content feature vector and the preference vector of the adjacent peer node ID2 exceeds the predetermined similarity range. In this case, the adjacent peer node ID2 is not selected as a content transmission destination. And peer node ID10 transmits a transmission confirmation request to the adjacent peer nodes of ID9 and ID16.
(S59) In contrast, the peer node ID 16 has already received the content from the peer node ID 17, and the sequence number has also been updated. Therefore, the peer node ID 16 returns a transmission confirmation response including the determination result NG to the peer node ID 10. On the other hand, the peer node ID 9 returns a transmission confirmation response including the determination result OK. As a result, the peer node 10 transmits the content only to the peer node ID 9.

  According to FIG. 6, it is assumed that the peer nodes 13 and 26 are suddenly down from the overlay network due to a power failure or the like. Even in such a case, by executing processing such as S58 and S59, the content is also distributed to the adjacent peer node ID 14 for the peer node 13 and the adjacent peer node 27 for the peer node 26. In this way, even when a peer node suddenly goes down, by applying the present invention, contents can be distributed while maintaining high reliability.

  FIG. 8 is a functional configuration diagram of the peer node in the present invention.

  The peer node 1 is connected to an overlay network virtually configured on a physical network connected by a plurality of peer nodes. According to FIG. 8, the peer node 1 has a communication interface unit 10 and a TCP / UDP / IP unit 11 on the network side.

  The peer node 1 also includes an overlay network layer unit 12, a content receiving unit 13, a content storage unit 14, a preference similarity determination unit 15, a content transmission unit 16, a first transmission destination selection unit 17, 2 transmission destination selection unit 18 and content distribution source node unit 19. These functional components are realized by executing a program that causes a computer mounted on the peer node to function.

  The overlay network layer unit 12 configures peer nodes in a multidimensional space (n dimensions: n ≧ 2), and executes, for example, a CAN protocol. According to FIG. 8, the overlay network layer unit 12 includes a routing table 121 and an adjacent peer confirmation unit 122.

  In the routing table 121, each peer node registers a preference vector of a user having an adjacent peer node that can communicate with one hop. An adjacent peer node that can communicate in one hop means a peer node in the front-rear direction for each dimension i (i = 1 to n). The identification value of the peer node is determined by the user preference vector, and participates in the CAN overlay network with the identification value.

  The adjacent peer confirmation unit 122 acquires a preference vector of the adjacent peer node by periodically transmitting a Keep-Alive message to the adjacent peer node registered in the routing table 121.

  The content receiving unit 13 receives content from a content distribution source node or an adjacent peer node in the i-dimensional direction. The received content is output to the content storage unit 14.

  The content storage unit 14 stores the content to be transferred and also stores the received content list. The received content list includes at least a content sequence number and a feature vector.

  The preference similarity determination unit 15 compares the feature amount vector of the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table, and determines whether the similarity is within a predetermined similarity range. judge.

  When it is determined that the similarity is within the predetermined similarity range, the content transmission unit 16 sends the content to the adjacent peer node selected by the first transmission destination selection unit 17 and the second transmission destination selection unit 18. Send.

  The first transmission destination selection unit 17 selects adjacent peer nodes in the 1 to (i-1) and (i + 1) dimensions as the transmission destination of the content. Here, the first transmission destination selection unit 17 includes a transmission confirmation request transmission unit 171, a transmission confirmation response reply unit 172, and a transmission determination unit 173. The transmission confirmation request transmission unit 171 transmits a transmission confirmation request including a sequence number to the adjacent peer node selected as the transmission destination of the content. When receiving the transmission confirmation request, the transmission confirmation response reply unit 172 compares the sequence number of the content already received with the sequence number included in the transmission confirmation request, and sends a transmission confirmation response including the comparison result to the peer node. Send back. When the transmission determination unit 173 receives the transmission confirmation response, the transmission determination unit 173 determines whether to transmit to the adjacent peer node based on the comparison result.

  For the i dimension, the second transmission destination selection unit 18 selects an adjacent peer node opposite to the content transmission source adjacent peer node as the transmission destination of the content.

  The content distribution source node unit 19 functions only when the peer node operates as a content distribution source node. The content distribution source node unit 19 includes a feature quantity vector extraction unit 191, a destination identification value derivation unit 192, a search request transmission unit 193, a search response reception unit 194, and a content distribution unit 195. The feature vector extraction unit 191 extracts a feature vector based on the content from the content to be transmitted using a category dictionary (see Table 4). The destination identification value deriving unit 192 derives a destination identification value based on the feature quantity vector (corresponding to the CAN identification value). The search request transmission unit 193 transmits a search request to the peer node closest to the destination identification value. The search response receiving unit 194 receives a search response. The content distribution unit 195 distributes the content toward the destination address included in the search response.

  When the peer node 1 is a personal computer, content distribution and browsing are operated by the user via the user interface. If the peer node is a set top box, the user interface may be a television.

  As described above in detail, according to the content distribution method, the peer node, and the program of the present invention, the overlay network is used and the content is determined while determining the similarity between the content feature vector and the user preference vector. To deliver. As a result, content can be distributed in real time only to users who are interested in the content without installing a specific content distribution server on the network.

  According to the present invention, since a specific content distribution server is not required, a cost increase that increases the number of servers according to the number of distribution users does not occur, and the scalability of the system can be ensured. Also, even if a peer node suddenly goes down and leaves the overlay network, the adjacent peer node is not affected in content distribution and the network bandwidth is not wasted.

  Various changes, modifications, and omissions of the above-described various embodiments of the present invention can be easily made by those skilled in the art. The above description is merely an example, and is not intended to be restrictive. The invention is limited only as defined in the following claims and the equivalents thereto.

DESCRIPTION OF SYMBOLS 1 Peanode, user terminal 10 Communication interface part 11 TCP / UDP / IP part 12 Overlay network layer part 121 Routing table 122 Adjacent peer confirmation part 13 Content receiving part 14 Content storage part 15 Preference similarity determination part 16 Content transmission part 17 1st Transmission destination selection unit 171 transmission confirmation request transmission unit 172 transmission confirmation response reply unit 173 transmission determination unit 18 second transmission destination selection unit 19 content distribution source node unit 191 feature vector extraction unit 192 destination identification value derivation unit 193 search request Transmission unit 194 Search response reception unit 195 Content distribution unit 2 Physical network 3 Overlay network

Claims (15)

A content distribution method in an overlay network virtually configured on a physical network connected by a plurality of peer nodes,
Each peer node has a routing table that registers a preference vector of a user having an adjacent peer node that can communicate in one hop,
Whether the peer node compares the feature vector based on the content to be transmitted with the preference vector of each neighboring peer node registered in the routing table, and whether or not the similarity is within a predetermined similarity range. A determination step for determining
A content distribution method comprising: a step of transmitting the content to the adjacent peer node based on the preference vector when it is determined that the similarity is within a predetermined similarity range. The overlay network is configured by arranging the peer nodes in a multidimensional space (n dimension: n ≧ 2),
The content distribution method according to claim 1, wherein the routing table registers the previous and next peer nodes as adjacent peer nodes for each dimension i (i = 1 to n).   The content delivery method according to claim 2, wherein the overlay network is a CAN (Content Addressable Network). About the previous stage of the determination step,
A receiving step in which the peer node receives the content from an adjacent peer node in an i-dimensional direction;
A first selection step of selecting an adjacent peer node in dimensions 1 to (i-1) and (i + 1) as a transmission destination of the content;
4. The content according to claim 2, further comprising a second selection step of selecting an adjacent peer node opposite to the content transmission source adjacent peer node as a transmission destination of the content for the i dimension. Delivery method. A sequence number is given to the data packet of the content to be transferred,
The first selection step is
The peer node transmitting a transmission confirmation request including the sequence number to an adjacent peer node selected as the transmission destination of the content;
When the adjacent peer node receives the transmission confirmation request, the adjacent peer node compares the sequence number of the content already received with the sequence number included in the transmission confirmation request, and sends a transmission confirmation response including the comparison result to the peer node. A step to reply,
5. The content distribution method according to claim 4, further comprising a step of determining whether or not to transmit to the adjacent peer node based on the comparison result when the peer node receives the transmission confirmation response. .   6. Each peer node acquires a preference vector of the neighboring peer node by periodically transmitting a Keep-Alive message to the neighboring peer node registered in the routing table. The content delivery method according to item. About the previous stage of the determination step,
A content distribution source node extracting from the content to be transmitted a feature vector based on the content;
The content distribution source node deriving a destination identification value based on the feature vector;
The content distribution source node sending a search request to a peer node of the destination identification value;
The content distribution source node includes a step of transmitting the content toward a destination address included in the search response when the search response is received. The content delivery method according to item. A peer node connected to a virtually configured overlay network on a physical network connected by a plurality of peer nodes,
A routing table in which preference vectors of users who have adjacent peer nodes that can communicate in one hop are registered;
A preference for comparing a feature vector based on the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table and determining whether the similarity is within a predetermined similarity range. Similarity determination means;
A peer node comprising: content transmission means for transmitting the content to the adjacent peer node based on the preference vector when it is determined that the similarity is within a predetermined similarity range. The overlay network is configured by arranging the peer nodes in a multidimensional space (n dimension: n ≧ 2),
The peer node according to claim 8, wherein the routing table registers the previous and next peer nodes as adjacent peer nodes for each dimension i (i = 1 to n).   The peer node according to claim 9, wherein the overlay network is a CAN. content receiving means for receiving the content from an adjacent peer node in the i-dimensional direction;
First transmission destination selection means for selecting adjacent peer nodes in the 1 to (i-1) and (i + 1) dimensions as transmission destinations of the content;
The i-dimension further comprises second transmission destination selection means for selecting an adjacent peer node opposite to the content transmission source adjacent peer node as a transmission destination of the content. The listed peer node. A sequence number is given to the data packet of the content to be transferred,
The first transmission destination selection means is
A transmission confirmation request transmission function for transmitting a transmission confirmation request including the sequence number to an adjacent peer node selected as a transmission destination of the content;
When the transmission confirmation request is received, the transmission confirmation response which compares the sequence number of the content already received with the sequence number included in the transmission confirmation request and returns a transmission confirmation response including the comparison result to the peer node Reply function,
The peer node according to claim 11, further comprising: a transmission determination function for determining whether to transmit to the adjacent peer node based on the comparison result when the transmission confirmation response is received.   13. The apparatus according to claim 8, further comprising an adjacent peer confirmation unit that acquires a preference vector of the adjacent peer node by periodically transmitting a Keep-Alive message to the adjacent peer node registered in the routing table. The peer node according to any one of the above. It has a content distribution source node means as a content distribution source,
The content distribution node means is
A feature vector extraction function for extracting a feature vector based on the content from the content to be transmitted;
A destination identification value derivation function for deriving a destination identification value based on the feature quantity vector;
A search request transmission function for transmitting a search request to a peer node of the destination identification value;
A search response receiving function for receiving a search response;
The peer node according to any one of claims 8 to 13, further comprising: a content distribution function for distributing the content toward a destination address included in the search response. A program for causing a computer mounted on a peer node connected to a virtually configured overlay network to function on a physical network connected by a plurality of peer nodes,
Each peer node has a routing table that registers preference vectors of users who possess neighboring peer nodes that can communicate in one hop, and
A preference for comparing a feature vector based on the content to be transmitted with the preference vector of each adjacent peer node registered in the routing table and determining whether the similarity is within a predetermined similarity range. Similarity determination means;
A program for a peer node, which causes a computer to function as a content transmission means for transmitting the content to the adjacent peer node based on the preference vector when it is determined that the similarity is within a predetermined similarity range. .

Images