Distributed information retrieval in peer-to-peer networks
AssigneeHewlett-Packard Development Company, L.P.
InventorsVana Kalogeraki, Dimitrios Gunopulos
Life of this patent
Estimating the term…
Abstract
A mechanism for information retrieval in fully decentralized, distributed, peer-to-peer network systems. Peer profiles are aggregated and collected in real-time by each peer. Each peer uses and integrates knowledge that it collects during query-reply cycles for each future query received, thereby learning over time and making information retrieval a more intelligent and rapid process. Each peer then autonomously decides which of its peers are most likely to have an answer to a given query. A routine is provided for monitoring the messages and profiling each peer, building a local peer profile for each node exchanging messages in the peer-to-peer network based on messages passing through the node.
Claims21
A method for conducting information retrieval in a distributed network of nodes, the method comprising:
- receiving, by a computing device, a current message;
- differentiating, by the computing device, the current message between a query message and a reply message, and
- if the current message is a reply message, associating, by the computing device, said reply message with a node address originating said reply message and storing data representative of said associating, said data including the node address originating the reply message and a topic of a query that resulted in the reply message;
- when said current message is a query, determining, by the computing device, a similarity factor from data representative of the current message to data representative of the prior received queries and ranking each known node with respect to the current message based on said similarity factor;
- forwarding, by the computing device, said current message to only nodes of the distributed network having a predetermined statistical probability of having a reply relevant to said query based on the similarity factor;
- and
- receiving, by the computing device, a reply from at least one of the nodes of the distributed network having a predetermined statistical probability of having a reply relevant to said query based on the similarity factor.
A computer readable medium containing a information retrieval tool for a peer-to-peer network, each peer's information retrieval tool comprising:
- input-output means for input and output of messages;
- coupled to said input-output means, search engine means for generating messages representative of informational queries and informational replies related to queries received;
- associated with the search engine means, 1) peer-profiler means for maintaining data representative of known peers query-reply history, 2) similarity means for determining a similarity of a current query to previous said queries received, the determining using the data representative of known peers query-reply history, and 3) associated with said peer profiler means and said similarity means, ranking means for ranking known peers with respect to the determined similarity of the current query;
- and
- a storage means for storing the similarity determined by the similarity means and the ranking of known peers with respect to the determined similarity of the current query.
A distributed network information retrieval system comprising:
- a network of autonomous peers, each having variable network connectivity, a temporary network address, and an informational database;
- a common networking protocol used by each of said peers;
- each peer according to said peer's proprietary system being able individually to analyze query messages received and to forward query messages only to a set of peers based on a derived peer ranking associated with likelihood of any particular peer having said informational database related to said informational retrieval query messages;
- and
- each peer being able to receive a reply from one of the set of peers to which the query messages are forwarded.
A computer-readable storage medium comprising software that, when executed by a processor, causes the processor to:
- receive network messages associated with information retrieval;
- differentiate the network messages between a query message and a reply message, and
- when a current message is a reply message, associate said reply message with a node address originating said reply message and storing data representative of said associating, said data including the node address originating the reply message and a topic of a query that resulted in the reply message;
- relate a current query message data to previously stored query message data by determining a similarity factor from data representative of the current message to data representative of the prior received queries;
- determine other nodes of the network having a predetermined statistical probability of having information related to the current query message from knowledge of previous reply messages issued by said other nodes with respect to said previously stored query message data and the similarity factor;
- and
- determine existence of local node information related to the current query message and for transmitting a reply message indicative of said determining to said query message.
A system comprising:
- a processor;
- and
- a program memory storing software that, when executed by the processor, causes the processor to:
- receive network messages associated with information retrieval;
- relate a current query message data to previously stored query message data by determining a similarity factor from data representative of the current message to data representative of the prior received queries;
- and
- determine other nodes of the network likely to have information related to the current query message based on knowledge of previous reply messages issued by said other nodes with respect to said previously stored query message data and the similarity factor;
- forward query messages only to a set of peers based on a derived peer node ranking associated with likelihood of any particular node having said informational database related to said query message;
- and
- receive a reply message from at least one of the set of peers to which the query messages are forwarded.
Dependencies are inferred from the claim text.