WO2010045859A1 - 内容节点双向聚类的系统、装置及方法 - Google Patents

内容节点双向聚类的系统、装置及方法 Download PDF

Info

Publication number
WO2010045859A1
WO2010045859A1 PCT/CN2009/074497 CN2009074497W WO2010045859A1 WO 2010045859 A1 WO2010045859 A1 WO 2010045859A1 CN 2009074497 W CN2009074497 W CN 2009074497W WO 2010045859 A1 WO2010045859 A1 WO 2010045859A1
Authority
WO
WIPO (PCT)
Prior art keywords
content
node
query
list
update
Prior art date
Application number
PCT/CN2009/074497
Other languages
English (en)
French (fr)
Inventor
李金龙
沈静波
刘姗姗
张进
王铁英
Original Assignee
华为技术有限公司
中国科学技术大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司, 中国科学技术大学 filed Critical 华为技术有限公司
Publication of WO2010045859A1 publication Critical patent/WO2010045859A1/zh

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00Network arrangements or protocols for supporting network services or applications
    • H04L67/01Protocols
    • H04L67/10Protocols in which an application is distributed across nodes in the network
    • H04L67/104Peer-to-peer [P2P] networks

Definitions

  • the present invention relates to the field of communications, and in particular, to a system, apparatus, and method for bidirectional clustering of content nodes.
  • Peer-to-Peer provides an efficient way to share content, find the content you need based on the key attributes of the content you need, and get a set of nodes with the content you need.
  • the number of nodes, the number of contents, and the number of copies of content in the network are also increasing, and the complexity of content search and copy management is also increasing.
  • the topology of distributed networks is optimized mainly by reducing the amount of traffic for content lookups and shortening the time required for content lookups.
  • P2P network structures such as Chord and Pasti based on DHT (Distributed Hash Table) structure
  • P2P networks based on hybrid structure such as SWOP (Small-world overlay P2P networks)
  • the routing table optimizes the topology and improves the efficiency of content lookup.
  • the semantic or ontology-based model first divides the content or interest by semantic or ontological methods, constructs a semantic network or an ontology tree, and predefines the association and edge between two points.
  • the weight either directly using an existing dictionary or ontology tree, and storing this knowledge on all nodes or in a location where all nodes can query.
  • each node can describe and associate the interest of the content and the node according to the knowledge, calculate the degree of association between the nodes, and then form clusters according to the interests.
  • this method can improve the efficiency of content search to a certain extent, especially the efficiency of the associated content search.
  • the behavior or historical information is used to analyze the behavior or history of the user, and the association between the content and the node is obtained, and clustering is formed on the basis of the cluster to accelerate the search.
  • a content node bidirectional clustering system provided by an embodiment of the present invention includes: a node layer and a content layer;
  • the node requesting the same content forms a loose cluster of nodes by querying the content by the node; in the content layer, the content requested by the same node forms a loose cluster of content according to the requirements of the node. .
  • the requesting node sends a first query request for the content to the node in the association list, so that the first query request is received
  • the first target node performs processing according to the first query request;
  • the association list stores Nm node information with the highest similarity to the content currently associated with the requesting node;
  • the requesting node receives the first target node and the feedback query is successful, establishing a download address list for the content of the query and downloading the content of the query;
  • the requesting node receives the first target node feedback query failure, sending a second query request to a node in its current neighbor that is closest to the first target node key value.
  • the association list of the requesting node is updated according to the calculated related content similarity.
  • the node that generates and publishes the content reviews the received update request, and determines to accept the update according to the audit result, and performs an update process
  • the node that generates and publishes the content publishes the update result and the new version number after accepting the update.
  • a requesting node configured to send a first query request to the first target node in the association list, and to the second target closest to the first target node key value when receiving the feedback of the first query request query failure Sending, by the node, a second query request, and establishing a download address list for the content of the query according to the feedback of the first or second target node, and downloading the content;
  • a target node including a first target node or a second target node, configured to receive the first or second query request, and determine, when the content of the query is satisfied according to the query request, to the requesting node Returns the available download address.
  • An obtaining unit configured to obtain a current content list of a node in the association list
  • a calculating unit configured to calculate, according to the content list acquired by the acquiring unit and the content list of the requesting node, the similarity degree of the related content of the requesting node and a node in the associated list;
  • an updating unit configured to update the association list of the requesting node according to the calculated related content similarity.
  • the query efficiency of the related content is improved by the first query, and the number of nodes that satisfy the query is increased by the second query;
  • the query efficiency of the content and ensure that at least one node that satisfies the query can be returned when the first query fails, so that even if the user's associated content set is dynamically changed, the node that meets the query condition can be quickly found.
  • it can be adapted to the current multi-content copy of the application environment, to ensure the consistency and validity of the copy under dynamically changing networks and constantly updated content.
  • FIG. 2 is a schematic diagram showing the relationship between associated nodes in a content in Embodiment 1 of the present invention
  • FIG. 3 is a schematic diagram of a data structure of an association list in Embodiment 1 of the present invention.
  • FIG. 4 is a schematic diagram of a data structure of a node in Embodiment 1 of the present invention.
  • FIG. 5 is a schematic flowchart of a method for querying content acquisition according to Embodiment 2 of the present invention.
  • FIG. 6 is a schematic diagram of a data structure of a query request in Embodiment 2 of the present invention.
  • FIG. 7 is a schematic flowchart of a method for updating an association list according to Embodiment 3 of the present invention.
  • FIG. 8 is a schematic flowchart of obtaining a current content list in Embodiment 3 of the present invention.
  • FIG. 9 is a schematic flow chart of a method for content update according to Embodiment 4 of the present invention.
  • FIG. 10 is a schematic structural diagram of a system for querying and acquiring content according to Embodiment 5 of the present invention.
  • FIG. 11 is a schematic structural diagram of an apparatus for updating an association list according to Embodiment 6 of the present invention.
  • the user's needs are directional.
  • the content that the user frequently needs for a period of time can be regarded as the related content set of the user at that moment, that is, the interest set of the node.
  • These content will have similarities or other associations with each other and can be considered as a loose clustering between content.
  • the same content is also cached by users who need it, so that there are many copies in the network. Since the users who store these copies request the same content, their associated content sets are intersected to varying degrees, so that associations between these users can be considered as loose clustering between nodes.
  • These two clusters respectively describe the association between the content and the association between the users, and the relationship between the content and the user establishes a connection between the two loose clusters.
  • the content is stored on the nodes, and the loose clustering between the contents objectively forms a loose cluster between the storage content nodes.
  • the association between these nodes is based on the similarity of the related content sets. Therefore, this association can be utilized to improve the efficiency of content lookup.
  • loose clustering between nodes is based on the need for the same content. This cluster can be used to manage the copy of the content, ensuring the update of the content and the copy under the lowest possible traffic. Effectiveness and consistency.
  • a content node bidirectional clustering system includes a node layer and a content layer.
  • nodes requesting the same content form a node loosely clustered by querying the content by the node;
  • content that is requested by the same node forms a loose cluster of content according to the requirements of the node.
  • nodes P1, P2, P3, P4, and P6 form a loose cluster of nodes for content A; and contents A, D, and E are loosely clustered of content formed by the requirements of node P4.
  • the CAN (Content-Addressable Network) route is used to construct the d-dimensional space between the nodes; and the content is allocated to the node with the closest key value according to the DHT mechanism.
  • Each content is mapped by DHT to a node with the closest key value. This node is called the CAN responsible node of the content.
  • the node PI in Figure 1 is the CP of content A.
  • the CP itself is not responsible for the management and update of the content, but as a backup node to store the information of the content, and as an index node when querying.
  • Each content may be needed by other nodes that access the content, store a copy of the content, and possibly manipulate or update the content.
  • These nodes are called replicas of the content (Replica Peer: RP), as shown In 1, the nodes P2, P3, P4, and P6 are all RPs of the content A.
  • the RP establishes a hierarchy to maintain and update content and replicas. RP is divided into The three types are: Virtual Server (VS), Consistent Replica Peer (CRP), and Inconsistent Replica Peer (IRP).
  • VS is the node that generates and publishes content, responsible for updating and publishing content.
  • the VS and CP maintain a two-way connection and periodic detection survives.
  • CRP is a node that is consistent with VS.
  • the copy of the content stored on CRP is kept up-to-date. After the content is updated,
  • the CRP will receive updates from the VS release.
  • the CRP maintains a bidirectional connection with the VS and maintains a timer. If there is no interaction information within a certain period, it will actively send information to detect the survival.
  • the IRP is a weakly consistent node with the VS.
  • the copy stored on the IRP may be an old version. After the content is updated, it will not receive the update information released by the VS. Before the IRP uses the content, you need to verify the version information with VS to maintain the consistency of the content.
  • the embodiment of the present invention establishes two loose clustering structures between the content and between the nodes, and establishes a connection between the two loose clusters by using the node's demand for the content. Thereby a bidirectional clustering topology of content nodes is formed. In this two-way loose clustering topology, nodes with similarly associated content sets establish connections with greater probability.
  • Embodiments of the present invention increase the range of capabilities and response speeds of content lookups by utilizing these connections, particularly the efficiency of querying associated content.
  • the embodiment of the present invention can also manage the copy by using loose clusters formed between nodes requesting the same content, and ensure the consistency and validity of the copy in the case of content update.
  • the relationship between the associated nodes of the same content is shown in Figure 2.
  • the node decides whether to become CRP or IRP according to its own requirements and performance.
  • the connection between CP and VS, VS and CRP is bidirectional, and the connection between VS and IRP is unidirectional, from IRP to VS.
  • the associated content set of these nodes contains this content, that is, their interests have similarities. .
  • the content stored on the node can be used as the associated content set of the node, and the content of the content stored by the two nodes is used to represent the current similar content similarity of the two nodes, namely:
  • Associative content similarity of nodes number of contents stored repeatedly by two nodes ( 1 )
  • Each node saves an associated list and stores the ⁇ nodes with the highest similarity to their current associated content.
  • the association list for each node, the identifier of the node, the IP address, the similarity of the associated content, the time when the similarity of the content associated with the node was last updated, the total number of contents owned by the node, and the most recent time are passed. The number of times the node successfully routes the recall result and the overall performance of the node.
  • the data structure of the associated list is shown in Figure 3.
  • the overall performance of the nodes in the association list is related to the node's current associated content similarity, the total number of content owned, and the number of times the query has been satisfied in the most recent period.
  • ⁇ ⁇ ⁇ ) ⁇ 5 ( ⁇ ⁇ ) + ⁇ ⁇ ( ⁇ ⁇ ) + ⁇ ⁇ ( ⁇ ⁇ ) m
  • Pf (P ⁇ ) is the comprehensive performance of the node ,, which is the similarity of the content of the node ⁇ and the node itself (available by the formula (1))
  • N D is the number of contents owned by the node Px
  • N D is the node Px The number of times the query was satisfied in the most recent period.
  • a, ⁇ is a manually set constant.
  • a common node has one of the following elements for the related content, such as CP, VS, CRP, and IRP.
  • the information stored in the common node can be as shown in FIG. 4, including: its own key value, key value space range, CAN routing table, demand content table, and associated list, where:
  • the CAN routing table is generated by the CAN protocol and records the key value and address information of all nodes adjacent to the logical node of the node.
  • the requirement content table records the key values of all the contents required by the node, the specific information of the content, and the CP where the content is located. And VS key value and address information;
  • the association list stores the related attributes of the N m nodes with the highest similarity to the content associated with itself.
  • Embodiment 2 of the present invention is a diagrammatic representation of Embodiment 2 of the present invention.
  • the node not only stores the information of the CAN neighbor node in the CAN routing table, but also stores the information of some distant nodes in the demand content list because the node needs the content, and also stores in the association list. Some nodes with high similarity to their own content.
  • Node queries for content can be divided into three categories:
  • a node queries a specified content
  • a node queries two interrelated content
  • the content node bidirectional clustering system according to the first embodiment is used, and the "binary query" is used to implement the searching and obtaining of the content.
  • the binary query includes: (1) The first query: that is, sending a "one-step” query in parallel to the nodes located in the associated list as an “association query” for accelerating the efficiency of the associated content query and increasing the satisfaction of the return.
  • the second query that is, by combining the remote connection and the CAN route as a "shortcut" in the small world model ", using "greedy query", improve the efficiency of ordinary content query, thus ensuring the efficiency of the first type of query, and to ensure that at least one node that satisfies the query can be returned when the associated query fails.
  • the associated query is concurrently sending query information to several nodes in the association list, and the node in the association list retrieves the content stored by itself after receiving the message: if there is content satisfying the query, if If not, the query information is directly discarded without further forwarding. Since the nodes in the association list and the query nodes themselves have similarities on the associated content set, the nodes in the association list will have the content satisfying the query with a higher probability, and the associated content can be accelerated by querying the nodes in the associated list. The query speed, and can return as many nodes as possible to satisfy the query request in a shorter time. When the associated query fails, the information query is not continuously forwarded, so that the query information can be prevented from exponentially growing in the network.
  • the greedy query is to send a greedy query request from the node information in the CAN routing table and the node information in the demand list, and select the node with the node key value closest to the target key value, and so on, until it finds that it has the content satisfying the query request. node.
  • This kind of search process is different from CAN's continuous approach to the destination in the key space, but by using the remote connection in the network as a "shortcut" jump approximation, you can quickly jump to the target within the first one or two hops. Near the node, and then find the target node by local precise positioning.
  • Step 51 The requesting node P A sends a first query request to an associated query request to the ⁇ ⁇ nodes with better comprehensive performance in the associated list.
  • the query request includes: a key value of the target content, and the request
  • a Boolean value can be used to identify whether a query request is an associated route or a greedy route.
  • the request node identifier and the random number generated by the requesting node may uniquely identify a query request.
  • Step 52 The node that receives the first query request performs processing according to the query request.
  • Each node has a request cache stack that is used to temporarily store query requests received in the most recent period.
  • the request cache stack records the request node identifier of the query request and the random number generated by the requesting node, and is replaced according to the first in first out FIFO principle.
  • step 52 specifically includes:
  • Step 521 After receiving a query request, a node first searches the request cache stack according to the request node identifier in the query request and the random number generated by the request node, and determines whether the query request has been received by itself: if it has received If yes, discard; otherwise, go to step 522.
  • Step 522 The node checks whether there is content queried by the node P A in the list of required content: if there is content satisfying the query, the process proceeds to step 523; otherwise, the process proceeds to step 524.
  • Step 523 The node has content that satisfies the query, and determines the information to be sent according to the identity of the content: a. If it is a CP or a CRP, the content itself is the latest version, and may directly The address information and the address of the VS of the content as the available download address, and carry the current content list of its own and send it to the node PA;
  • the VS address of the content is used as the available download address, and the current content list is sent to the node PA.
  • Step 524 If the node does not satisfy the content of the query, the query information is directly discarded, and the feedback query fails.
  • Step 53 If the node PA receives the feedback of the successful query, establish a temporary download address list for the content of the query and download the content.
  • the node P A After receiving the feedback information of other nodes, the node P A stores the available download address in the feedback information into the download address list. The addresses already included in the download address list are not stored repeatedly. Thereafter, the node P A establishes a connection using the address in the download address list to start the download.
  • the feedback that the node P A receives the query success refers to the available download address of the queried content that receives the feedback.
  • CP, VS, or CRP returned to the requesting node a query message as the feedback node will automatically piggyback a list of content itself, if the CP, VS, or CRP is not associated with the request in the list of nodes P A, P A is the requesting node calculates a feedback Node comprehensive performance, if the comprehensive performance of the feedback node is better than the node with the worst comprehensive performance in the current association list, replace the node with the worst comprehensive performance with the feedback node.
  • Step 54 if the node receiving the query failure P A feedback, the requesting node key P A neighbor will issue a second key from the target of the current query request greedy nearest node a query request.
  • the query fails to be fed back to the node P A , and the node P A will re-initiate the second query request to ensure that the first query request fails. At least one node that satisfies the query.
  • Step 55 The node that receives the query request performs processing according to the query request.
  • step 55 specifically includes:
  • Step 551 After receiving a query request, a node first searches the request cache stack according to the request node identifier in the query request and the random number generated by the request node, and determines whether the query request has been received by itself: if it has received If yes, discard; otherwise, go to step 552.
  • Step 552 The node checks whether there is content queried by the node P A in the list of required content: if there is content satisfying the query, the process proceeds to step 553; otherwise, the process proceeds to step 554.
  • Step 553 The node has content that satisfies the query, and determines the information to be sent according to the identity of the content: a. If it is a CP or a CRP, the content itself is the latest version, and may directly The address information and the address of the VS of the content as the available download address, and carry the current content list of its own and send it to the node P A;
  • the VS address of the content is used as the available download address, and the current content list is carried to the node P A.
  • Step 554 The node does not satisfy the content of the query, and further forwards according to the principle that the key value is closest.
  • Step 56 The node P A receives the result of the second query request, and establishes a temporary download address list for the content of the query. After receiving the feedback information of other nodes, the node P A stores the available download address in the feedback information into the download address list. The addresses already included in the download address list are not stored repeatedly. Thereafter, the node P A establishes a connection using the address in the download address list to start the download.
  • CP, VS, or CRP returned to the requesting node a query message as the feedback node will automatically piggyback a list of content itself, if the CP, VS, or CRP is not associated with the request in the list of nodes P A, P A is the requesting node calculates a feedback Node comprehensive performance, if the comprehensive performance of the feedback node is better than the node with the worst comprehensive performance in the current association list, replace the node with the worst comprehensive performance with the feedback node.
  • the query capability of the content is improved, and the foregoing types 1, 2, and 3 queries can be simultaneously solved, for example:
  • a node queries two interrelated contents: For example, node P1 queries two interrelated contents A and B. When the node P1 queries the first content A, the node P1 obtains the content A through the node P2. If the content lists of nodes P1 and P2 have a higher similarity, node P1 will add node P2 to the associated list. Looking for and content A at node P1 When the content B is associated with each other, since the content A and the content B are associated with each other, the node P2 is likely to simultaneously store the contents A and 8. The node P1 can directly obtain the content B by only one hop by sending an association query request to P2 in the association list.
  • nodes P1 and P2 both query content A. After the content P is queried by the node P1, the content A is cached in the content list of the node P1.
  • the nodes P1 and P2 may be nodes associated with each other, possibly in the association list of each other, the node P2 sends an association query request to the node P1. , that is, content A can be obtained directly by only one hop.
  • the embodiment may return a part of the node that satisfies the query request after the first hop, thereby starting the download; at this time, the second in the binary query
  • the second query request continues to be routed through the network until the VS of the content is found, thereby returning the VS and all CRP nodes, increasing the source node that can be downloaded. That is to say, the downloading of the content of this embodiment is divided into two steps: First, a small number of nodes are found in the first hop to start downloading, and then the download source is added by the second query request.
  • This embodiment can start downloading at a faster speed, thereby reducing the waiting time of the user. This difference is especially noticeable when there are many nodes in the network. For example, when there are 10000 nodes in the network, the average query path length of the CAN model is 16.49, which is 2.63 in this embodiment. That is to say, the CAN model can return to the first node that satisfies the query after an average of 16.49 hops, and this embodiment can return a small number of nodes after the first hop to start downloading, and after 2.63 hops, further Increase the download source.
  • the reaction speed of this embodiment is 13.49 hops faster than CAN, which can greatly reduce the waiting time of the user.
  • the ability of the first hop in this embodiment to return a part of the nodes to start downloading does not change, and the ability of the present embodiment to improve the query response speed is even more obvious.
  • the embodiment Since the embodiment sends ⁇ ⁇ associated query messages concurrently in the first query of the binary query, the ⁇ ⁇ traffic is only sent from the requesting node to the node in the associated list, and is not in the Forwarding in the network does not increase the traffic of the transit node, but only increases the traffic of the client.
  • the client's request is not a frequent process, but rather sparse. It takes less traffic for each query and does not affect the performance of the client. Therefore, the increased traffic of the first query does not adversely affect the performance of the embodiment and is negligible.
  • Embodiment 3 of the present invention is a diagrammatic representation of Embodiment 3 of the present invention.
  • the node in the association list should maintain a high degree of similarity of the associated content with the requesting node itself. Since the requirements of the nodes are dynamically changing, the content stored on the nodes is also dynamically changing. In addition, the content on the network changes as the nodes join and leave, so the similarity between the nodes is also changing.
  • the association list needs to be updated so that the current requirements of the node can be accurately described, so that nodes can form better clusters according to requirements.
  • the method for updating the association list in this embodiment includes:
  • Step 71 Obtain a current content list of a node in the association list.
  • the content list of the user is actively taken.
  • the content list of the attached node itself is selected based on the existing message in the network. , you can minimize the increased traffic.
  • the content list includes only the identifier of the content stored on one node, and does not include the specific content information.
  • the requesting node when it sends the query information, it does not carry its own content list for two reasons: (1) Since the query request needs to be forwarded several times to reach the target node, if the content list is piggybacked, Increase the traffic of all forwarding nodes; (2) Since the requesting node does not have the information of the required content at this time, the content list will change after the request information is returned, so the content list at this time is not accurate.
  • Step 81 setting the associated T content similarity is valid, a node detects an association list in ⁇ ⁇ node (e.g., PB) last update time interval exceeds the period T from the present time;
  • ⁇ ⁇ node e.g., PB
  • Step 82 the node P A sends a request to the node P B to query the current content list of the node P B ;
  • step 83 the node P A receives the list of its own content returned by the node P B.
  • the latest content list of a node in the associated list can be periodically detected and acquired.
  • the requesting node can obtain the current content list of a node in the associated list.
  • Step 72 The requesting node calculates its own content similarity with the node
  • Step 73 The requesting node updates its own association list according to the calculated similarity of the related content.
  • the requesting node updates the related attribute of the node (for example, PB) in the association list according to the calculated similar content similarity.
  • the associated content set of the node and the similarity of the related content between the computing nodes are described according to the content currently stored by the node, thereby avoiding the storage space consumed by the large knowledge base required by the semantic and ontological methods.
  • Embodiment 4 of the present invention is a diagrammatic representation of Embodiment 4 of the present invention.
  • this embodiment implements control of the replica by loosely clustering the nodes where the content and its replica are located. And updates.
  • the node that generates and publishes the content as the content of the VS is responsible for the update and release of the content.
  • the CP maintains a strong two-way connection with the VS.
  • the other nodes request the content, they choose the CRP or IRP that becomes the content according to their own needs and performance. If the query node needs to access the content frequently, and its performance is good (for example, the bandwidth is large, the processing speed is fast, etc.), you can choose to become CRP, and establish a two-way connection with the VS to maintain strong consistency.
  • VS, CP, CRP, and IRP form a loose cluster for the same content through these connections, and based on this cluster, different content updates and update release modes are used according to the type of node to reduce traffic.
  • each content has a version number, and the VS of the content operates on the version number after each legal update.
  • the method for updating the content of this embodiment includes:
  • Step 91 A node requesting content update (for example, CRP or IRP) sends an update request to a node (for example, VS) that generates and distributes content; Specifically, when an RP wants to operate and update the content, it first determines its type: If it is CRP, the node itself stores the latest version of the content, can directly operate the content, and send updates to the VS. Request, the request includes: an operation result and a current version number;
  • the copy stored on the node may be an old version.
  • the current version number needs to be sent to the VS to detect the version information. If the version number is the same as the current version stored on the VS, the VS returns a confirmation message. Inconsistent, VS returns the content information of the current latest version. The IRP then operates on the content in the same way as CRP and sends an update request to the VS.
  • Step 92 The VS reviews the received update request and processes the result according to the audit result
  • the VS After receiving the update request, the VS first determines whether the version number of the content is the latest version number. If yes, accepts the update and modifies the version number, and returns a confirmation message; otherwise, rejects the update and returns the latest version information. If two RPs submit the update results at the same time, the VS determines the validity of the update based on the timestamp of the update request and accepts the earlier update of the timestamp.
  • Step 93 VS releases the update result and the new version number after the update
  • the update result and the new version number are sent to the CP and all CRPs. Since the IRP and VS are weakly consistent, there is no need to release update information to the IRP.
  • VS, CP, CRP, and IRP form a loose cluster of nodes for the same content through a bidirectional or unidirectional connection between them, and different types of nodes are used according to the type of the node.
  • Content update and update release mode to reduce traffic.
  • by actively releasing updates to all CRPs a large amount of inquiry information brought by CRP frequent access to the content is avoided; the latest version of the content is actively inquired to the VS before the content is used by the IRP, thereby avoiding a low frequency of accessing the content.
  • IRP frequently releases the traffic brought by the update.
  • this embodiment Based on loose clustering of this node, this embodiment implements management and control of content and its copy under lower traffic, ensuring that all nodes can obtain the freshest version when accessing content, and ensure that All update operations are cumulatively incremented based on the latest version, and there is no update to the old version. It also effectively reduces system traffic and adapts to the needs of dynamically changing environments.
  • Embodiment 5 of the present invention is a diagrammatic representation of Embodiment 5 of the present invention.
  • a fifth embodiment of the present invention provides a system for querying content, including:
  • the requesting node 101 is configured to send a first query request to the node in the association list, and send a second to the node whose key value is closest to the target node key value when receiving the feedback that the first query request query fails Querying the request, and establishing a temporary download address list for the content of the query according to the content list fed back by the target node, and downloading the content;
  • the target node 102 is configured to receive the first or second query request, and perform processing according to the query request.
  • target node 102 includes:
  • the first determining unit 1021 is configured to: according to the request node identifier in the query request and the random number generated by the requesting node, retrieve the request cache stack, determine whether the target node has received the query request, and if it has been received, discard
  • the second requesting unit 1022 is configured to: when the first determining unit 1021 determines that the target node has not received the query request, further determine whether the content of the query is in the target content content list of the target node;
  • the first control unit 1023 is configured to, when the second determining unit 1022 determines that the content of the target content in the target content list has content that satisfies the query, determine the information to be sent according to the identity of the target node with respect to the content;
  • the first The control unit 1023 may directly use the address information of the target node and the address of the content VS as an available download address, and carry the current content list of the target node to the requesting node;
  • the first control unit 1023 directly uses the address information of the target node and all the CRP addresses of the stored content as available download addresses, and carries the current content list of the target node to the requesting node;
  • the first control unit 1023 uses the VS address of the content as an available download address, and carries the current content list of the target node to the requesting node.
  • the second control unit 1024 is configured to perform different processing according to the type of the query when the second determining unit 1022 determines that the content of the target node does not satisfy the content of the query:
  • the query information is directly discarded, and no processing is performed;
  • Embodiment 6 of the present invention is a diagrammatic representation of Embodiment 6 of the present invention.
  • an embodiment of the present invention provides an apparatus for updating an association list, including:
  • the obtaining unit 111 is configured to obtain a current content list of a node in the association list.
  • the calculating unit 112 is configured to calculate an association content similarity between the requesting node and a node in the association list;
  • the updating unit 113 is configured to update the association list of the requesting node according to the calculated related content similarity.
  • the obtaining unit 111 may obtain the current content list fed back by a node in the association list by querying the process of acquiring the content.
  • the content list may also be obtained by periodic timing detection of a node in the association list.
  • the method includes the steps of the foregoing method embodiments; and the foregoing storage medium includes: a medium that can store program codes, such as a ROM, a RAM, a magnetic disk, or an optical disk.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Information Transfer Between Computers (AREA)

Description

内容节点双向聚类的系统、 装置及方法
本申请要求于 2008年 10月 24日提交中国专利局、 申请号为 200810218623.0、 发明名称为"内 容节点双向聚类的系统、装置及方法"的中国专利申请的优先权, 其全部内容通过引用结合在本申请 中。 技术领域
本发明涉及通信领域, 尤其涉及一种内容节点双向聚类的系统、 装置及方法。
发明背景
对等网络 P2P (Peer-to-Peer)提供了一种分布式内容共享的有效途径, 可以根据所需内容的关 键属性, 找到所需的内容, 并可以获得拥有所需内容的一组节点。 但随着用户的不断增加, 网络中 的节点数、 内容数以及内容的副本数也在不断增加, 内容査找和副本管理的复杂度也在加大。 目前 主要通过降低内容査找的通信量, 缩短内容査找所需的时间, 来优化分布式网络的拓扑结构。 比如 基于 DHT (Distributed Hash Table, 分布式哈希表) 结构的 Chord、 Pasti 等 P2P网络结构, 还有基 于混合结构的 P2P网络, 如 SWOP ( Small-world overlay P2P networks ) , 这些模型都通过构建额外的 路由表优化拓扑, 提高内容査找的效率。
例如, 现有方案一中, 基于语义或本体论的模型首先通过语义或本体论的方式对内容或兴趣进 行划分, 构建成一个语义网络或一个本体树, 预定义两点之间的关联和边的权重, 或者直接使用已 有的词典或本体树, 并将这些知识存储在所有的节点上或放置在一个所有节点都可以査询的位置。 这样每个节点都可以根据这些知识对内容和节点的兴趣进行描述和关联,并计算节点之间的关联度, 然后根据兴趣形成聚类。 在这种聚类结构中, 由于关联的内容一般位于相同的聚类之内, 所以这种 方式一定程度可以提高内容査找的效率, 特别是关联内容査找的效率。
又如, 现有方案二中, 基于行为或历史信息的方式使用用户的行为或历史记录进行分析, 获取 内容和节点之间的关联, 在此基础上形成聚类, 加速査找。
在实现本发明的过程中, 发明人发现现有技术至少存在以下缺陷: 现有的方案虽然在一定程度 上提高了内容査找的效率, 但并没有充分利用用户之间以及内容之间的固有关联, 在动态变化的网 络和不断更新的内容下, 内容的査找效率依然不高, 也无法适应多副本的应用环境。 发明内容
有鉴于此, 有必要提供一种内容节点双向聚类的系统、 装置及方法, 以便在当前网络规模不断 增加的情况下, 仍能进行高效的内容査询、 关联列表更新或内容更新。
本发明实施例提供的一种内容节点双向聚类系统, 包括: 节点层和内容层;
所述节点层中, 请求同一内容的节点之间通过节点对内容的査询形成节点松散聚类; 所述内容层中, 被同一个节点请求的内容之间根据节点的需求形成内容松散聚类。
本发明实施例提供的一种利用上述系统査询获取内容的方法, 包括:
请求节点向关联列表中的节点发送对某内容的第一査询请求, 以使得接收所述第一査询请求的 第一目标节点根据所述第一査询请求进行处理; 所述关联列表中存储有和所述请求节点当前关联内 容相似度最高的 Nm个节点信息;
如果所述请求节点接收到所述第一目标节点反馈査询成功, 则为所述査询的内容建立下载地址 列表并下载所述査询的内容;
如果所述请求节点接收到所述第一目标节点反馈査询失败, 则向自身当前邻居中距离所述第一 目标节点键值最近的节点发送第二査询请求。
本发明实施例提供的一种利用上述系统进行关联列表更新的方法, 包括:
获取关联列表中某节点当前的内容列表; 所述关联列表中存储有和所述请求节点当前关联内容 相似度最高的 Nm个节点信息;
根据所述获取的内容列表和所述请求节点的内容列表计算请求节点与所述关联列表中某节点的 关联内容相似度;
根据计算得到的关联内容相似度更新请求节点的关联列表。
本发明实施例提供的一种利用上述系统进行内容更新的方法, 包括:
接收请求内容更新的节点向产生和发布内容的节点发送的更新请求;
所述产生和发布内容的节点对接收的更新请求进行审核, 并根据审核结果确定接受更新后, 进 行更新处理;
所述产生和发布内容的节点在接受更新后发布更新结果和新的版本号。
本发明实施例提供的一种査询获取内容的系统, 包括:
请求节点, 用于向关联列表中的第一目标节点发送第一査询请求, 在接收到所述第一査询请求 査询失败的反馈时向距离第一目标节点键值最近的第二目标节点发送第二査询请求, 以及根据所述 第一或第二目标节点的反馈为査询的内容建立下载地址列表并下载所述内容;
目标节点, 包括第一目标节点或第二目标节点, 分别用于接收所述第一或第二査询请求, 并根 据所述査询请求确定拥有满足査询的内容时, 向所述请求节点返回可用下载地址。
本发明实施例提供的一种关联列表更新的装置, 包括:
获取单元, 用于获取关联列表中某节点当前的内容列表;
计算单元, 用于根据所述获取单元获取的内容列表和请求节点的内容列表计算请求节点与所述 关联列表中某节点的关联内容相似度;
更新单元, 用于根据计算得到的关联内容相似度更新请求节点的关联列表。
通过本发明实施例提供的技术方案, 在当前网络规模不断增加的情况下, 通过第一査询提高关 联内容的査询效率, 增加返回的满足査询的节点数; 通过第二査询提高普通内容的査询效率, 并保 证在第一査询失败时至少能返回一个满足査询的节点, 这样即使用户的关联内容集处于动态变化之 中, 也能够迅速的找到符合査询条件的节点。 此外, 能够适合当前多内容副本的应用环境, 能在动 态变化的网络和不断更新的内容下, 保证副本的一致性和有效性。 附图简要说明 图 1是本发明实施例一一种内容节点双向聚类的系统的原理示意图;
图 2是本发明实施例一中一个内容的关联节点之间关系示意图;
图 3是本发明实施例一中关联列表的数据结构示意图;
图 4是本发明实施例一中节点的数据结构示意图;
图 5是本发明实施例二一种査询获取内容的方法的具体流程示意图;
图 6是本发明实施例二中査询请求的数据结构示意图;
图 7是本发明实施例三一种关联列表更新的方法的流程示意图;
图 8是本发明实施例三中获取当前内容列表的具体流程示意图;
图 9是本发明实施例四一种内容更新的方法的流程示意图;
图 10是本发明实施例五一种査询获取内容的系统的结构示意图;
图 11是本发明实施例六一种关联列表更新的装置的结构示意图。
实施本发明的方式
在实际网络中, 用户的需求是有趋向性的, 用户在一段时间内频繁需求的内容可以视为用户在 该时刻的关联内容集, 也就是节点的兴趣集。 这些内容彼此之间会存在相似性或其他的关联, 可以 视为一个内容间的松散聚类。 而同一内容也会被需求它的用户缓存, 从而在网络中拥有很多副本。 存储了这些副本的用户由于请求了同一内容, 所以他们的关联内容集在不同程度上有所交叉, 从而 在这些用户之间形成了关联, 可以视为一个节点间的松散聚类。 这两种聚类分别描述了内容之间的 关联和用户之间的关联, 而内容和用户之间的需求关系则在这两种松散聚类之间建立了连接。
内容存储在节点之上, 内容之间的松散聚类客观上形成了存储内容节点之间的松散聚类, 这些 节点之间的关联是建立在关联内容集相似的基础上。 因此可以利用这种关联提高内容査找的效率。 另一方面, 节点间的松散聚类建立在对同一内容的需求之上, 这个聚类可以被用来对内容的副本进 行管理, 在尽量低的通信量之下保证内容的更新发布以及副本的有效性和一致性。
以下结合附图对本发明实施例进行详细描述。
如图 1所示, 本发明实施例一内容节点双向聚类系统包括节点层和内容层, 所述节点层中, 请 求同一内容的节点之间通过节点对内容的査询形成节点松散聚类; 所述内容层中, 被同一个节点请 求的内容之间根据节点的需求形成内容松散聚类。 例如, 节点 Pl, P2, P3 , P4, P6形成了针对内 容 A的节点松散聚类; 而内容 A, D, E则是由节点 P4的需求而形成的内容松散聚类。
节点之间采用 CAN (Content- Addressable Network, 内容寻址网络)路由构建了 d维空间; 而内 容则根据 DHT机制分配到键值最接近的节点之上。
每个内容通过 DHT 映射到一个键值最接近的节点上, 这个节点称为该内容的 CAN负责节点
(CAN Peer: CP), 如图 1中的节点 PI是内容 A的 CP。 CP本身并不负责内容的管理和更新, 而是 作为一个备份节点保存内容的信息, 并作为査询时的索引节点。
每个内容可能会被其他节点需要, 这些节点对内容进行访问, 存储了内容的副本, 并可能对内 容进行操作或更新, 这些节点称为该内容的复制节点 (Replica Peer: RP), 如图 1中, 节点 P2、 P3、 P4、 P6都是内容 A的 RP。 RP之间通过建立层次结构完成对内容以及副本的维护和更新。 RP分为 三类: 虚拟服务器(Virtual Server: VS)、 强一致节点 (Consistent Replica Peer: CRP), 弱一致节点 ( Inconsistent Replica Peer: IRP )。
VS是产生和发布内容的节点, 负责对内容进行更新并发布。 VS和 CP保持双向连接, 并周期 性的探测存活。
CRP是和 VS保持强一致的节点, CRP上存储的内容副本一直保持最新的版本,在内容更新后,
CRP会接收到 VS发布的更新信息。 CRP和 VS之间保持双向连接, 并保持定时器, 在一定周期内 若没有交互信息, 则会主动发送信息探测存活。
IRP是和 VS保持弱一致的节点, IRP上存储的副本可能是陈旧的版本, 在内容发生更新后, 不 会接收到 VS发布的更新信息。在 IRP使用内容前,需要先向 VS验证版本信息来保持内容的一致性。
本发明实施例根据节点对内容的需求以及历史信息, 在内容之间以及节点之间建立了两种松散 聚类结构, 并利用节点对内容的需求在两种松散聚类之间建立了连接, 从而形成了内容节点的双向 聚类拓扑结构。 在这个双向松散聚类拓扑下, 具有相似关联内容集的节点之间以较大的概率建立连 接。 本发明实施例通过利用这些连接提高了内容査找的能力范围和响应速度, 特别是关联内容的査 询效率。 此外, 本发明实施例还可通过利用请求同一个内容的节点之间形成的松散聚类对副本进行 管理, 在内容更新的情况下保证了副本的一致性和有效性。
同一内容的关联节点之间的关系如图 2所示, 节点根据自身的需求和性能决定是成为 CRP还是 IRP。 CP和 VS之间、 VS和 CRP之间的连接是双向的, VS和 IRP之间的连接是单向的, 从 IRP指 向 VS。 此外, 网络中还随机存在着若干在査询路由时建立的由 IRP或 CRP发起的指向 CRP的单向 连接。 这些连接都是远程连接。
由于 VS、 CRP, IRP这三种节点存储内容或其副本这个行为都是建立在节点本身对这个内容的 需求之上, 所以这些节点的关联内容集包含了这个内容, 即它们的兴趣具有相似性。
可以使用节点上存储的内容作为节点的关联内容集, 用两个节点共同存储的内容数表示这两个 节点当前的关联内容相似度, 即:
节点的关联内容相似度=两个节点重复存储的内容个数 ( 1 )
每个节点保存一个关联列表,存储了和自己当前关联内容相似度最高的 ^^个节点。在关联列表 中, 对于每个节点都记载了该节点的标识符、 IP地址、 关联内容相似度、 该节点关联内容相似度最 近一次更新的时间、 该节点拥有的内容总数、 最近一段时间通过该节点进行路由成功召回结果的次 数以及节点的综合性能。 关联列表的数据结构如图 3所示。
关联列表中节点的综合性能和节点当前的关联内容相似度、 拥有的内容总数、 以及最近一段时 间满足査询的次数相关。 节点当前的关联内容相似度越高、 拥有的内容数越多、 最近一段时间满足 査询的次数越多, 则该节点在以后的査询中能够返回所需内容的可能性越大, 则该节点当前的综合 性能越好。
可以使用下列公式来计算节点的综合性能:
^χ)=αΝ5χ) + βΝτχ) +χΝκχ) m 其中: Pf(P^)是节点 Ρχ的综合性能, 是节点 Ρχ和节点自身的关联内容相似度 (由公 式( 1 )可得), ND是节点 Px拥有的内容数, ND是节点 Px在最近一段时间满足査询的次 数。 a、 β、 是人工设置的常量。 应当理解, 如果调整上述三个因素 (节点关联内容相似度、 拥 有的内容数, 以及最近一段时间满足査询的次数)之间的权重比例, 或选择其中一到两个因素使用, 作为综合性能计算公式, 原理仍是相同, 仍然属于本发明的保护范围。
对任意普通节点而言, 只要它所负责的键值空间内有内容存在, 它即作为 CP 负责这些内容。 当它需要其他内容并需要持续保持这些内容时, 它将成为内容的 RP。所以, 一个普通节点对于相关 内容有且仅有 CP、 VS、 CRP、 IRP这些身份中的一个。
普通节点中存储的信息可以如图 4所示, 包括: 自身的键值、 键值空间范围、 CAN路由表、 需 求内容表和关联列表, 其中:
CAN路由表由 CAN协议产生, 记录所有与本节点逻辑邻接的节点的键值和地址信息; 需求内容表记录了该节点所需的所有内容的键值, 内容的具体信息, 该内容所在的 CP和 VS的 键值和地址信息;
关联列表存储了和自己关联内容相似度最高的 Nm个节点的相关属性。
本发明实施例二:
由实施例一可知, 节点不仅在 CAN路由表中存储了 CAN邻居节点的信息, 还因为节点对内容 的需求而在需求内容列表中存储了一些远处节点的信息, 此外在关联列表中还存储了一些和自身关 联内容相似度较高的节点。
节点对内容的査询可以分为 3类:
1、 一个节点査询一个指定内容;
2、 一个节点査询两个相互关联的内容;
3、 两个节点査询同一个内容;
本实施例利用实施例一所述的内容节点双向聚类系统, 采用"二元査询"实现对内容的査找和获 取。 二元査询包括: (1 )第一査询: 即向位于关联列表中的节点并行的发送"一步"査询作为"关联査 询",用于加速关联内容査询效率,增加返回的满足査询的节点数,从而以较高的概率实现了第 2类、 第 3类査询; (2)第二査询: 即通过将远程连接和 CAN路由相结合作为小世界模型中的"捷径", 采 用"贪心査询", 提高普通内容的査询效率, 从而保证了第 1类査询的高效性, 并保证在关联査询失 败时至少能返回一个满足査询的节点。
具体的, 关联査询是并发的向关联列表中的若干个节点发送査询信息, 关联列表中的节点在接 收到消息之后会检索自身存储的内容: 如果有满足査询的内容则返回; 如果没有, 则直接丢弃査询 信息, 而不会进一步转发。 由于关联列表中的节点和査询节点自身在关联内容集上具有相似性, 所 以关联列表中的节点会以较高概率拥有满足査询的内容, 通过査询关联列表中的节点可以加速关联 内容的査询速度, 并可以在较短的时间内返回尽量多的满足査询请求的节点。而在关联査询失败时, 不继续转发信息査询, 则可以避免査询信息在网络中呈指数增长。 此外, 由于关联列表中节点的关 联列表所存储的节点和査询节点的关联内容相似度会进一步降低, 再继续转发査询信息, 成功的概 率只会进一步降低, 所以关联査询信息在査询失败的情况下不会被转发。
贪心査询是从 CAN路由表中的节点信息以及需求列表中的节点信息中,选择节点键值距离目标 键值最近的节点发送贪心査询请求, 如此反复, 直到找到拥有满足査询请求内容的节点。 这种査找 过程不同于 CAN在键值空间连续的逼近目的地,而是通过利用网络中的远程连接作为"捷径"的跳跃 式逼近, 可以在最初一两跳 (hop) 内迅速地跳到目标节点附近, 然后再通过局部的精确定位找到目 标节点。
请参照图 5所示, 本实施例査询获取内容的方法的具体流程如下:
步骤 51, 请求节点 PA向关联列表中综合性能较好的 Νκ个节点发送第一査询请求一关联査询 请求, 如图 6所示, 査询请求包括: 目标内容的键值, 请求节点标识符, 请求节点生成的随机数, 请求节点 IP地址, 关联路由和贪心路由的标识位。可以通过一个布尔值标识一个査询请求是关联路 由还是贪心路由。 其中, 请求节点标识符和请求节点生成的随机数可以唯一标识一个査询请求。
步骤 52, 接收所述第一査询请求的节点, 根据所述査询请求进行处理。
每个节点拥有一个请求缓存栈, 用于暂存最近一段时间接收到的査询请求。 该请求缓存栈记录 了査询请求的请求节点标识符和请求节点生成的随机数, 根据先进先出 FIFO原则替换。
进一步的, 步骤 52具体包括:
步骤 521, 当一个节点接收到査询请求之后, 首先根据査询请求中的请求节点标识符以及请求 节点生成的随机数检索请求缓存栈, 判断自身是否已经接收过这个査询请求: 如果已经接收过, 则 丢弃; 否则, 转入步骤 522。
步骤 522, 所述节点査看自身的需求内容列表中是否有节点 PA所査询的内容: 如果有满足査询 的内容, 则转入步骤 523 ; 否则转入步骤 524。
步骤 523, 所述节点拥有满足査询的内容, 根据自身相对于所述内容的身份决定发送的信息: a、 如果是 CP或 CRP, 则自身存储的内容本身就是最新的版本, 可以直接将自身的地址信息以 及所述内容的 VS的地址作为可用下载地址, 并携带上自身当前的内容列表发送给节点 PA;
b、 如果是 VS, 则直接将自身的地址信息以及存储的所述内容的所有 CRP地址作为可用下载地 址, 并携带上自身当前的内容列表发送给节点 PA;
c 如果是 IRP, 由于自身存储的内容可能不是最新的版本, 则将所述内容的 VS地址作为可用 下载地址, 并携带上自身当前的内容列表发送给节点 PA。
步骤 524, 所述节点没有满足査询的内容, 则直接丢弃査询信息, 反馈査询失败。
步骤 53, 如果节点 PA接收到査询成功的反馈, 则为所述査询的内容建立一个临时的下载地址 列表并下载内容。
节点 PA在接收到其他节点的反馈信息后, 将反馈信息中的可用下载地址存储到下载地址列表 中。 下载地址列表中已经包含的地址, 不会重复存储。 此后, 节点 PA使用下载地址列表中的地址建 立连接, 开始下载。本实施例中节点 PA接收到査询成功的反馈即是指接收到反馈的所査询内容的可 用下载地址。 CP、 VS或 CRP作为反馈节点向请求节点返回査询信息时, 会主动捎带自身的内容列表, 如果 CP、 VS或 CRP并不在请求节点 PA的关联列表中, 则请求节点 PA将计算反馈节点综合性能, 如果 反馈节点的综合性能优于当前关联列表中综合性能最差的节点, 则用反馈节点替换该综合性能最差 的节点。
步骤 54, 如果节点 PA接收到査询失败的反馈, 则请求节点 PA将向当前邻居中键值距离目标键 值最近的节点发出第二査询请求一贪心査询请求。
例如上述步骤 524中, 该节点没有满足査询的内容, 则向节点 PA反馈査询失败, 此时节点 PA 将再发起第二査询请求, 以保障第一査询请求失败时能返回至少一个满足査询的节点。
步骤 55, 接收所述査询请求的节点, 根据所述査询请求进行处理。
与步骤 52类似, 步骤 55具体包括:
步骤 551, 当一个节点接收到査询请求之后, 首先根据査询请求中的请求节点标识符以及请求 节点生成的随机数检索请求缓存栈, 判断自身是否已经接收过这个査询请求: 如果已经接收过, 则 丢弃; 否则, 转入步骤 552。
步骤 552, 所述节点査看自身的需求内容列表中是否有节点 PA所査询的内容: 如果有满足査询 的内容, 则转入步骤 553 ; 否则转入步骤 554。
步骤 553, 所述节点拥有满足査询的内容, 根据自身相对于所述内容的身份决定发送的信息: a、 如果是 CP或 CRP, 则自身存储的内容本身就是最新的版本, 可以直接将自身的地址信息以 及所述内容的 VS的地址作为可用下载地址, 并携带上自身当前的内容列表发送给节点 PA;
b、 如果是 VS, 则直接将自身的地址信息以及存储的所述内容的所有 CRP地址作为可用下载地 址, 并携带上自身当前的内容列表发送给节点 PA;
c、 如果是 IRP, 由于自身存储的内容可能不是最新的版本, 则将所述内容的 VS地址作为可用 下载地址, 并携带上自身当前的内容列表发送给节点 PA
步骤 554, 所述节点没有满足査询的内容, 则根据键值最近的原则进一步转发。
步骤 56, 节点 PA接收第二査询请求的结果, 为所述査询的内容建立一个临时的下载地址列表。 节点 PA在接收到其他节点的反馈信息后, 将反馈信息中的可用下载地址存储到下载地址列表 中。 下载地址列表中已经包含的地址, 不会重复存储。 此后, 节点 PA使用下载地址列表中的地址建 立连接, 开始下载。
CP、 VS或 CRP作为反馈节点向请求节点返回査询信息时, 会主动捎带自身的内容列表, 如果 CP、 VS或 CRP并不在请求节点 PA的关联列表中, 则请求节点 PA将计算反馈节点综合性能, 如果 反馈节点的综合性能优于当前关联列表中综合性能最差的节点, 则用反馈节点替换该综合性能最差 的节点。
本实施例提高了内容的査询能力, 可以同时解决前述第 1、 2、 3三类査询, 例如:
对于第 2类査询, 一个节点査询两个相互关联的内容: 如节点 P1査询两个相互关联的内容 A 和 B。 在节点 P1査询第一个内容 A时, 节点 P1通过节点 P2获得了内容 A。 如果节点 P1和 P2的 内容列表具有较高的相似性,则节点 P1会将节点 P2添加到关联列表中。在节点 P1要寻找和内容 A 相互关联的内容 B时, 由于内容 A和内容 B相互关联, 所以节点 P2很可能同时存储内容 A和8。 节点 P1通过向关联列表中的 P2发送关联査询请求, 就可以仅通过一个 hop就直接获得内容 B。
对于第 3类査询, 两个节点査询同一个内容: 如节点 P1和 P2都査询内容 A。在节点 P1査询了 内容 A后, 内容 A会缓存在节点 P1的内容列表中。 当节点 P2査询内容 A时, 由于节点 P1和 P2 请求了同一个内容, 节点 P1和 P2可能是相互关联的节点, 可能在彼此的关联列表中, 节点 P2通过 向节点 P1发送关联査询请求, 即可以仅通过一个 hop就直接获得内容 A。
通过本实施例的二元査询中的第一査询, 本实施例可以在第 1个 hop之后就返回一部分满足査 询请求的节点, 从而开始下载; 此时, 二元査询中的第二査询请求继续在网络中路由, 直到找到内 容的 VS, 从而返回 VS以及所有的 CRP节点, 增加可以下载的源节点。 也就是说, 本实施例内容 的下载分为两步: 先在第 1个 hop找到少量的节点开始下载, 再通过第二査询请求增加下载源。
本实施例可以以更快的速度开始下载, 从而减少用户的等待时间。 特别是当网络中的节点数很 多的时候, 这种差异更加明显。 比如网络中有 10000个节点时, CAN模型的平均査询路径长度是 16.49, 本实施例是 2.63。 也就是说, CAN模型要在平均 16.49跳之后才能返回第一个满足査询的节 点, 而本实施例在第 1个 hop之后就可以返回一小部分节点开始下载, 而在 2.63跳之后, 进一步增 加下载源。 本实施例的反应速度比 CAN快了 13.49跳, 可以大幅降低用户的等待时间。
并且, 随着网络中节点数目的增加, 本实施例这种第 1跳就能够返回一部分节点开始下载的能 力并不会发生变化, 而且, 本实施例可以提高査询响应速度的能力反而会更加明显。
由于本实施例在二元査询的第一査询中并发地发送了 Νκ个关联査询消息,但这 Νκ个通信量只是 从请求节点发送到关联列表中的节点, 并不会在网络中转发, 不会增加中转节点的通信量, 而是仅 仅增加了客户端的通信量。 而客户端的请求并不是一个频繁的过程, 而是相对稀疏的, 对于每次査 询多承担一点通信量并不会对客户端的性能产生影响。 所以, 第一査询所增加的通信量并不会对本 实施例的性能带来负面的影响, 是可以忽略的。
本发明实施例三:
为了保证实施例二中第一査询能够以较高的概率获得满足査询请求的节点, 关联列表中的节点 应该和请求节点自身保持较高的关联内容相似度。 由于节点的需求处于动态变化之中, 节点上存储 的内容也在动态变化, 此外网络上的内容也随着节点的加入离开而不断变化, 所以节点之间的关联 内容相似度也是处于变化中的, 需要对关联列表进行更新, 以便能够精确的描述节点当前的需求, 使得节点之间能根据需求形成较好的聚类。
请参照图 7所示, 本实施例关联列表更新的方法, 包括:
步骤 71, 获取关联列表中某节点当前的内容列表;
具体的获取方式有两种:
1、 在例如实施例二的査询内容的流程中获取;
由实施例二可知,在 CP、 VS或 CRP向请求节点返回査询信息时,会主动捎带自身的内容列表, 这种在网络中现有消息的基础上有选择的附带上节点自身的内容列表,可以尽量减少增加的通信量。
由于只有 CP、 VS或 CRP有最新版本的内容信息, 所以当请求节点 PA发送査询请求之后, 经过若 干次转发, 査询信息最终会落到这三种节点上。 这三种节点在返回内容信息时, 会捎带上自身的内 容列表。 内容列表仅包括一个节点上存储内容的标识符, 不包括具体内容信息。 此处, 在请求节点 发送査询信息时, 并不会捎带自身的内容列表, 原因有两点: (1 ) 由于査询请求需要经过若干次转 发才能到达目标节点, 如果捎带了内容列表, 会增加所有转发节点的通信量; (2 ) 由于请求节点此 时还没有所需内容的信息, 内容列表在请求信息返回之后还会发生变化, 所以此时的内容列表并不 准确。
2、 定时获取; 请参照图 8所示, 具体流程如下:
步骤 81, 设置关联内容相似度的有效期为 T, 检测到节点 ΡΑ的关联列表中的某一节点 (例如 PB) 最近一次更新时间距离当前时间的间隔超过了有效期 T;
步骤 82, 节点 PA向节点 PB发送询问节点 PB当前内容列表的请求;
步骤 83, 节点 PA接收节点 PB返回的其自身当前的内容列表。
这种方式可以周期性的定时检测并获取关联列表中的某一节点最新的内容列表。
通过上述两种方式, 请求节点即可获取关联列表中某节点当前的内容列表。
步骤 72, 请求节点计算自身与所述节点的关联内容相似度;
步骤 73, 请求节点根据计算得到的关联内容相似度更新自身关联列表。
具体的, 请求节点将根据计算得到的关联内容相似度更新自身关联列表中所述节点 (例如 PB ) 的相关属性。
本实施例中, 根据节点当前存储的内容描述节点的关联内容集以及计算节点之间的关联内容相 似度, 从而避免了语义和本体论的方式所需的庞大知识库所消耗的存储空间。
本发明实施例四:
由于所有内容的 RP上都会保存一份该内容的副本, 所以整个网络中会存在内容的多个副本。 为了保证所有节点都能在较低的通信量下, 获得新鲜有效的内容并能对内容进行合法的操作, 本实 施例通过对内容及其副本所在节点进行松散节点聚类, 实现对副本的控制和更新。
产生和发布内容的节点作为内容的 VS 负责内容的更新和发布。 CP作为索引和备份节点, 和 VS 建立双向连接保持强一致。 其他节点在请求内容之后, 根据自身的需求和性能选择成为内容的 CRP或 IRP。 如果査询节点需要经常访问该内容, 且自身性能较好 (例如: 带宽较大, 处理速度较 快等), 则可以选择成为 CRP, 和 VS建立双向连接保持强一致。 如果査询节点并不需要经常访问该 内容, 或自身性能并不太好, 也可以选择成为 IRP, 和 VS建立单向连接保持弱一致。 VS、 CP、 CRP 以及 IRP通过这些连接形成针对同一内容的一个松散聚类, 并在这个聚类的基础上根据节点的类型 采用不同的内容更新和更新发布模式, 以降低通信量。
为了能在内容更新之后区分内容的不同版本, 每个内容都有一个版本号, 由该内容的 VS在每 次合法更新之后对版本号进行操作。
请参照图 9所示, 本实施例内容更新的方法, 包括:
步骤 91, 请求内容更新的节点 (例如 CRP或 IRP ) 向产生和发布内容的节点 (例如 VS ) 发送 更新请求; 具体的, 当一个 RP要对内容进行操作及更新时, 首先会判断自身的类型: 如果是 CRP, 则该节点本身就存储了内容的最新版本, 可以直接对内容进行操作, 并向 VS发 送更新请求, 该请求包括: 操作结果和当前的版本号;
如果是 IRP,贝 IJ该节点上存储的副本可能是陈旧版本, 需要先向 VS发送当前版本号来检测版本 信息, 如果该版本号和 VS上存储的当前版本一致, 则 VS返回确认信息, 如果不一致, 则 VS返回 当前最新版本的内容信息。此后 IRP采用和 CRP同样的方式对内容进行操作并向 VS发送更新请求。
步骤 92, VS对接收的更新请求进行审核, 并根据审核结果进行处理;
VS接收到更新请求后, 先判断内容的版本号是否是最新的版本号, 如果是, 则接受更新并修改 版本号, 同时返回确认信息; 否则, 拒绝更新, 并返回最新版本信息。 如果有两个 RP 同时提交更 新结果, VS根据更新请求的时间戳判定更新的有效性, 接受时间戳较早的更新。
步骤 93, VS在更新后发布更新结果和新的版本号;
在 VS接受更新之后, 会将更新结果和新的版本号发送给 CP和所有 CRP。 由于 IRP和 VS之间 是弱一致, 所以并不需要向 IRP发布更新信息。
本实施例中, VS、 CP、 CRP以及 IRP通过它们之间的双向或单向的连接, 形成了针对同一内 容的节点松散聚类, 并在这个聚类的基础上根据节点的类型采用不同的内容更新和更新发布模式, 以降低通信量。 本实施例通过主动向所有 CRP发布更新, 避免了 CRP频繁访问内容所带来的大量 询问信息; 通过 IRP在使用内容前主动向 VS询问内容的最新版本, 从而避免了向访问内容频率较 低的 IRP频繁发布更新所带来的通信量。 基于这个节点松散聚类, 本实施例在较低的通信量下实现 了对内容及其副本的管理和控制, 保证了所有节点在对内容进行访问时都能获得最新鲜的版本, 并 保证了所有更新操作都是在最新版本的基础上累计递增的, 而不会出现对陈旧版本的更新操作。 此 外还有效地降低了系统通信量, 并能适应动态变化环境的需求。
本发明实施例五:
请参照图 10所示, 本发明实施例五提供一种査询获取内容的系统, 包括:
请求节点 101, 用于向关联列表中的节点发送第一査询请求、 在接收到所述第一査询请求査询 失败的反馈时, 向键值距离目标节点键值最近的节点发送第二査询请求, 以及根据目标节点反馈的 内容列表为査询的内容建立临时的下载地址列表, 并下载所述内容;
目标节点 102, 用于接收所述第一或第二査询请求, 并根据所述査询请求进行处理。
进一步的, 所述目标节点 102包括:
第一判断单元 1021, 用于根据査询请求中的请求节点标识符以及请求节点生成的随机数检索请 求缓存栈, 判断目标节点是否已经接收过所述査询请求, 如果已经接收过, 则丢弃所述査询请求; 第二判断单元 1022, 用于当第一判断单元 1021判断目标节点未接收过所述査询请求时, 进一步 判断目标节点需求内容列表中是否有所述査询的内容;
第一控制单元 1023, 用于当所述第二判断单元 1022判断目标节点的需求内容列表中有满足査 询的内容时, 根据目标节点相对于所述内容的身份决定发送的信息;
具体的, 如果目标节点是 CP或 CRP, 则目标节点自身存储的内容本身就是最新的版本, 第一 控制单元 1023可以直接将目标节点的地址信息以及所述内容 VS的地址作为可用下载地址, 并携带 上目标节点当前的内容列表发送给请求节点;
如果目标节点是 VS, 则第一控制单元 1023直接将目标节点的地址信息以及存储的所述内容的 所有 CRP地址作为可用下载地址, 并携带上目标节点当前的内容列表发送给请求节点;
如果目标节点是 IRP, 则第一控制单元 1023将所述内容的 VS地址作为可用下载地址, 并携带 上目标节点当前的内容列表发送给请求节点。
第二控制单元 1024, 用于当所述第二判断单元 1022判断目标节点的需求内容列表中没有满足 査询的内容时, 根据査询的类型进行不同的处理:
如果是第一査询, 则直接丢弃査询信息, 不做任何处理;
如果是第二査询, 则根据键值最近的原则进一步转发。
本发明实施例六:
请参照图 11所示, 本发明实施例六提供一种关联列表更新的装置, 包括:
获取单元 111, 用于获取关联列表中某节点当前的内容列表;
计算单元 112, 用于计算请求节点与所述关联列表中某节点的关联内容相似度;
更新单元 113, 用于根据计算得到的关联内容相似度更新请求节点的关联列表。
进一步的, 所述获取单元 111可以通过査询获取内容的流程获取关联列表中某节点反馈的当前 的内容列表; 也可以通过对关联列表中某节点周期性的定时检测获取所述内容列表。
本领域普通技术人员可以理解: 实现上述方法实施例的全部或部分步骤可以通过程序指令相关 的硬件来完成, 前述的程序可以存储于一计算机可读取存储介质中, 该程序在执行时, 执行包括上 述方法实施例的步骤; 而前述的存储介质包括: ROM, RAM, 磁碟或者光盘等各种可以存储程序代码 的介质。
以上所述, 仅为本发明的具体实施方式, 但本发明的保护范围并不局限于此, 任何熟悉本技术 领域的技术人员在本发明揭露的技术范围内, 可轻易想到的变化或替换, 都应涵盖在本发明的保护 范围之内。 因此, 本发明的保护范围应该以权利要求的保护范围为准。

Claims

权利要求
1、 一种内容节点双向聚类系统, 其特征在于: 包括节点层和内容层;
所述节点层中, 请求同一内容的节点之间通过节点对内容的査询形成节点松散聚类; 所述内容层中, 被同一个节点请求的内容之间根据节点的需求形成内容松散聚类。
2、 根据权利要求 1所述的系统, 其特征在于: 根据对某内容的需求, 所述节点层的节点包括: 内容寻址网络 CAN负责节点 CP, 用于作为备份节点保存所述内容的信息, 并作为査询所述内 容时的索引节点;
复制节点 RP, 用于存储所述内容的副本, 和 /或对所述内容进行操作或更新。
3、 根据权利要求 2所述的系统, 其特征在于: 所述复制节点 RP包括:
虚拟服务器 VS, 所述 VS和所述 CP双向连接, 用于产生和发布所述内容, 负责对所述内容进行 更新并发布;
强一致节点 CRP, 与所述 VS双向连接, 当所述 VS更新内容时, 接收所述 VS发布的更新信息; 弱一致节点 IRP, 与所述 VS形成由所述 IRP指向所述 VS的单向连接, 在使用内容前, 先向所述 VS验证自身存储的内容的版本信息以保持内容的一致性。
4、 根据权利要求 1所述的系统, 其特征在于: 所述节点存储的信息包括:
自身的键值、 键值空间范围、 CAN路由表、 需求内容表和 /或关联列表。
5、 根据权利要求 1所述的系统, 其特征在于: 所述关联列表记载的信息包括:
所述节点的标识符、 IP地址、 关联内容相似度、 所述关联内容相似度最近一次更新的时间、 所 述节点拥有的内容总数、 最近一段时间通过所述节点进行路由成功召回结果的次数和 /或所述节点的 综合性能。
6、 一种利用如权利要求 1-5任一项所述的系统査询获取内容的方法, 包括:
请求节点向关联列表中的节点发送对某内容的第一査询请求, 以使得接收所述第一査询请求的 第一目标节点根据所述第一査询请求进行处理; 所述关联列表中存储有和所述请求节点当前关联内 容相似度最高的 Nm个节点信息;
如果所述请求节点接收到所述第一目标节点反馈査询成功, 则为所述査询的内容建立下载地址 列表并下载所述査询的内容;
如果所述请求节点接收到所述第一目标节点反馈査询失败, 则向自身当前邻居中距离所述第一 目标节点键值最近的节点发送第二査询请求。
7、 根据权利要求 6所述的方法, 其特征在于: 所述接收所述第一査询请求的节点根据所述第一 査询请求进行处理, 具体包括:
所述第一目标节点根据所述第一査询请求中的请求节点标识符以及请求节点生成的随机数检索 请求缓存栈, 判断自身是否已经接收过所述第一査询请求, 如果已经接收过, 则丢弃所述第一査询 请求;
如果判断自身未接收过所述第一査询请求, 则所述第一目标节点进一步査看自身的需求内容列 表中是否有所述请求节点所査询的内容; 如果所述第一目标节点拥有满足査询的内容, 则向所述请求节点返回可用下载地址; 如果所述第一目标节点没有满足査询的内容, 则直接丢弃査询信息, 反馈査询失败。
8、 根据权利要求 7所述的方法, 其特征在于: 所述第一目标节点拥有满足査询的内容时, 还向 所述请求节点返回第一目标节点当前的内容列表, 且所述返回操作具体包括:
如果所述第一目标节点是 CP或 CRP,则可以直接将自身的地址信息以及所述内容的 VS的地址 作为可用下载地址, 并携带上自身当前的内容列表发送给所述请求节点;
如果所述第一目标节点是 VS, 则直接将自身的地址信息以及存储的所述内容的所有 CRP地址 作为可用下载地址, 并携带上自身当前的内容列表发送给所述请求节点;
如果所述第一目标节点是 IRP,则将所述内容的 VS地址作为可用下载地址,并携带上自身当前 的内容列表发送给所述请求节点。
9、 根据权利要求 6所述的方法, 其特征在于: 还包括:
所述第二目标节点根据所述第二査询请求中的请求节点标识符以及请求节点生成的随机数检索 请求缓存栈, 判断自身是否已经接收过所述査询请求, 如果已经接收过, 则丢弃所述第二査询请求; 如果判断自身未接收过所述第二査询请求, 则所述第二目标节点进一步査看自身的需求内容列 表中是否有所述请求节点所査询的内容;
如果所述第二目标节点拥有满足査询的内容, 则向所述请求节点返回可用下载地址; 如果所述第二目标节点没有满足査询的内容, 则根据键值最近原则进一步转发所述第二査询请 求。
10、 根据权利要求 9所述的方法, 其特征在于: 所述第二目标节点拥有满足査询的内容时, 还 向所述请求节点返回第二目标节点当前的内容列表所述返回操作具体包括:
如果所述第二目标节点是 CP或 CRP,则可以直接将自身的地址信息以及所述内容 VS的地址作 为可用下载地址, 并携带上自身当前的内容列表发送给所述请求节点;
如果所述第二目标节点是 VS, 则直接将自身的地址信息以及存储的所述内容所有 CRP地址作 为可用下载地址, 并携带上自身当前的内容列表发送给所述请求节点;
如果所述第二目标节点是 IRP,则将所述内容的 VS地址作为可用下载地址,并携带上自身当前 的内容列表发送给所述请求节点;
11、 一种利用如权利要求 1-5任一项所述的系统进行关联列表更新的方法, 包括:
获取关联列表中某节点当前的内容列表; 所述关联列表中存储有和所述请求节点当前关联内容 相似度最高的 Nm个节点信息;
根据所述获取的内容列表和所述请求节点的内容列表计算请求节点与所述关联列表中某节点的 关联内容相似度;
根据计算得到的关联内容相似度更新请求节点的关联列表。
12、 根据权利要求 11所述的方法, 其特征在于: 所述获取关联列表中某节点当前的内容列表, 具体包括:
通过査询获取内容的流程获取关联列表中某节点反馈的当前的内容列表; 和 /或通过对关联列表 中某节点周期性的定时检测获取所述某节点当前的内容列表。
13、 一种利用如权利要求 1-5任一项所述的系统进行内容更新的方法, 包括:
接收请求内容更新的节点向产生和发布内容的节点发送的更新请求;
所述产生和发布内容的节点对接收的更新请求进行审核, 并根据审核结果确定接受更新后, 进 行更新处理;
所述产生和发布内容的节点在接受更新后发布更新结果和新的版本号。
14、 根据权利要求 13所述的方法, 其特征在于: 所述请求内容更新的节点向产生和发布内容的 节点发送更新请求, 具体包括:
所述请求内容更新的节点判断自身的类型:
如果是 CRP, 则直接对内容进行操作, 并向所述产生和发布内容的节点发送更新请求, 所述请 求包括: 操作结果和当前的版本号;
如果是 IRP, 则先向所述产生和发布内容的节点发送当前版本号来检测版本信息, 如果该版本 号和所述产生和发布内容的节点上存储的当前版本一致, 则所述产生和发布内容的节点返回确认信 息, 如果不一致, 则所述产生和发布内容的节点返回当前最新版本的内容信息; 此后所述 IRP向所 述产生和发布内容的节点发送更新请求。
15、 根据权利要求 13所述的方法, 其特征在于: 所述产生和发布内容的节点对接收的更新请求 进行审核, 并根据审核结果确定接受更新后, 进行更新处理, 具体包括:
所述产生和发布内容的节点判断内容的版本号是否是最新的版本号, 如果是, 则接受更新并修 改版本号, 同时返回确认信息; 否则, 拒绝更新, 并返回最新版本信息。
16、 根据权利要求 13所述的方法, 其特征在于: 如果有两个同时提交的、 针对同一内容的更新 请求, 则所述产生和发布内容的节点根据所述更新请求的时间戳来判定更新的有效性, 接受时间戳 较早的更新。
17、 根据权利要求 13所述的方法, 其特征在于: 所述产生和发布内容的节点在接受更新后发布 更新结果和新的版本号, 具体包括:
所述产生和发布内容的节点接受更新之后, 将更新结果和新的版本号发送给 CP和所有 CRP。
18、 一种査询获取内容的系统, 其特征在于: 包括:
请求节点, 用于向关联列表中的第一目标节点发送第一査询请求, 在接收到所述第一査询请求 査询失败的反馈时向距离第一目标节点键值最近的第二目标节点发送第二査询请求, 以及根据所述 第一或第二目标节点的反馈为査询的内容建立下载地址列表并下载所述内容;
目标节点, 包括第一目标节点或第二目标节点, 分别用于接收所述第一或第二査询请求, 并根 据所述査询请求确定拥有满足査询的内容时, 向所述请求节点返回可用下载地址。
19、 根据权利要求 18所述的系统, 其特征在于: 所述目标节点进一步包括:
第一判断单元, 用于根据第一査询请求或第二査询请求中的请求节点标识符以及请求节点生成 的随机数检索请求缓存栈, 判断所述目标节点是否已经接收过所述査询请求, 如果已经接收过, 则 丢弃所述査询请求; 第二判断单元, 用于当所述第一判断单元判断目标节点未接收过所述査询请求时, 进一步判断 目标节点需求内容列表中是否有所述査询的内容;
第一控制单元,用于当所述第二判断单元判断目标节点的需求内容列表中有满足査询的内容时, 向所述请求节点返回可用下载地址;
第二控制单元, 用于当所述第二判断单元判断目标节点的需求内容列表中没有满足査询的内容 时, 在所述査询请求为第二査询请求的情况下, 根据键值最近的原则转发所述第二査询请求。
20、 一种关联列表更新的装置, 其特征在于: 包括:
获取单元, 用于获取关联列表中某节点当前的内容列表;
计算单元, 用于根据所述获取单元获取的内容列表和请求节点的内容列表计算请求节点与所述 关联列表中某节点的关联内容相似度;
更新单元, 用于根据计算得到的关联内容相似度更新请求节点的关联列表。
21、 根据权利要求 20所述的装置, 其特征在于: 所述获取单元通过査询获取内容的流程获取关 联列表中某节点反馈的当前的内容列表; 和 /或通过对关联列表中某节点周期性的定时检测获取所述 某节点当前的内容列表。
PCT/CN2009/074497 2008-10-24 2009-10-19 内容节点双向聚类的系统、装置及方法 WO2010045859A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN200810218623.0 2008-10-24
CN 200810218623 CN101729387A (zh) 2008-10-24 2008-10-24 内容节点双向聚类的系统、装置及方法

Publications (1)

Publication Number Publication Date
WO2010045859A1 true WO2010045859A1 (zh) 2010-04-29

Family

ID=42118956

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2009/074497 WO2010045859A1 (zh) 2008-10-24 2009-10-19 内容节点双向聚类的系统、装置及方法

Country Status (2)

Country Link
CN (1) CN101729387A (zh)
WO (1) WO2010045859A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103533050A (zh) * 2013-10-15 2014-01-22 珠海金山网络游戏科技有限公司 一种消除服务器缓存的方法及系统
CN105429796B (zh) * 2015-11-12 2019-03-29 鲁东大学 一种基于智能控制器的远距离无线转发方法
CN110932972B (zh) * 2020-01-23 2020-06-05 飞诺门阵(北京)科技有限公司 一种数据传输方法、装置及电子设备
US11706097B2 (en) 2020-01-23 2023-07-18 Novnet Computing System Tech Co., Ltd. Task processing method applied to network topology, electronic device and storage medium

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101110733A (zh) * 2007-08-21 2008-01-23 北京航空航天大学 大规模动态异构混合无线自组织网络中的接入及路由计算方法
CN101594316A (zh) * 2008-05-30 2009-12-02 华为技术有限公司 一种分布式网络的管理方法、内容查询方法、系统及装置

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101110733A (zh) * 2007-08-21 2008-01-23 北京航空航天大学 大规模动态异构混合无线自组织网络中的接入及路由计算方法
CN101594316A (zh) * 2008-05-30 2009-12-02 华为技术有限公司 一种分布式网络的管理方法、内容查询方法、系统及装置

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
"Proceedings of the Fourth International Conference on Peer-to-Peer computing (P2P'04), 27 Aug. 2004(27.08.2004)", article XIAOLE BAI ET AL.: "ICN: Interest-based Clustering Network" *

Also Published As

Publication number Publication date
CN101729387A (zh) 2010-06-09

Similar Documents

Publication Publication Date Title
US5933849A (en) Scalable distributed caching system and method
WO2011150830A1 (zh) 获取内容的方法、节点及内容网络
JP5090450B2 (ja) 階層に編成され、ネットワークを介してリンクされた複数のノードに保管された複製データを更新するための方法、プログラム、およびコンピュータ可読媒体
US10193805B2 (en) User oriented IoT data discovery and retrieval in ICN networks
US20050108368A1 (en) Method and apparatus for representing data available in a peer-to-peer network using bloom-filters
WO2010118638A1 (zh) 一种基于地理位置信息编码的cdn网络系统及数据分发方法
JP5541812B2 (ja) コンテンツを公表するための方法及びシステム、コンテンツを検索するための方法及びシステム
WO2013010432A1 (zh) 一种对等网络中数据存储和查询的方法、节点及系统
WO2009143738A1 (zh) 一种分布式网络的管理方法、内容查询方法、系统及装置
CN106326308A (zh) 一种基于sdn的网内重复数据删除方法及系统
WO2010045859A1 (zh) 内容节点双向聚类的系统、装置及方法
Li et al. SCOM: A scalable content centric network architecture with mobility support
US20140222988A1 (en) Method for adaptive content discovery for distributed shared caching system
TWI466502B (zh) 訊息傳送路由安排平臺
Clay et al. Posting protocol for improved keyword search success in peer-to-peer file-sharing systems
Steiner et al. Evaluating and improving the content access in KAD
WO2014075426A1 (zh) 一种数据更新方法及web服务器
CN113901144B (zh) 非全网共识区块链下的查询方法、设备及存储介质
US11218841B2 (en) Fast query method for dynamic location information of mobile network entity
CN112256638A (zh) Cnfs协议中有限去中心化分布式哈希表资源查找方法
JP2010182301A (ja) 自己組織型分散オーバーレイ・ネットワークにおいてオブジェクトへの参照を分散させる方法、コンピュータプログラム、及びノード、並びに自己組織型分散オーバーレイ・ネットワーク
Ma et al. A resource discovery mechanism integrating p2p and grid
Tran et al. Multi-keyword based information routing in peer-to-peer networks
Ghassan et al. SEMANTIC BASED INFORMATION-CENTRIC NETWORKINGOUTING ALGORITHMS
Dakshayini et al. Hierarchical cooperative-peers fog caching strategy to improve the performance of content centric networking

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 09821580

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 09821580

Country of ref document: EP

Kind code of ref document: A1