Font Size: a A A

Information Retrieval Based On Peer-to-Peer Computing

Posted on:2005-02-21Degree:DoctorType:Dissertation
Country:ChinaCandidate:B LingFull Text:PDF
GTID:1118360125967530Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Peer-to-peer computing (Abbr. P2P) has become a hot spot in the computerresearch and industrial fields. In a P2P-based system, each peer (node) has equalfunctionalities and responsibilities, i.e., a peer can both act as a server of othersproviding them with data/services, and serve as a client of others consuming theirdata/services; Furthermore, the interaction among peers can be direct and symmetric;In addition, peers can join in or leave from the system at any time to form dynamicnetwork environment. This sort of systems enjoy many potential merits, includingscalability, reliability, high information availability, and efficiently utilizing systems'resources, so that P2P is regarded as a promising technology to re-architecture thefuture Internet-based applications. While file sharing has become the most popular research and development topicin the P2P computing field, existing systems can just support semantics-free sharingof large granularity and inefficiently utilize their own resources. To address thesechallenges, we take the text files as sharing objects and propose the concept ofpeer-to-peer based information retrieval system. Furthermore, we have conducted anextensive study to this new type of systems to address the challenges related to its keytechniques and achieved following contributions: (â…°) We have proposed a three-layer based architecture, which is made up of structured layer, unstructured layer and application layer. This new type of architecture has inherited the advantages of both structured and unstructured architectures while eliminated their respective disadvantages as well, so that the potential merits of peer-to-peer computing can be exploited to efficiently support information retrieval in the dynamic P2P environment. (â…±) Based on the comprehensive analysis on the existing resource location and query routing schemes, we have proposed a peer-clustering based resource location mechanism and a self-adaptive routing strategy, which can efficiently utilize systems' resources and effectively satisfy users' demands. (â…²) We have systematically studied the challenges related to ranking and merging the answers retrieved from different peers in a P2P-based information system (e.g. PeerIS), and have obtained a deep insight to its underlying diathesis. Furthermore, we have proposed a fully distributed ranking and merging strategy for P2P-based information retrieval systems 6Abstract and addressed relevant issues. (â…³) We have conducted an analysis on retrieving optimization and peer dynamics for the P2P-based information retrieval systems. Specifically, we have proposed a high level cost model for this sort of systems and divised an agent-based strategy to obtain the coefficients of the model. Furthermore, with taking the dynamics of peers as a factor of P2P systems' cost, we have proposed a fuzzy cost analysis on peers' dynamics, i.e., we have employed the fuzzy theory to depict and capture the peers' behavior and identify their reliability. Consequently, the goal of retrieving optimization has been defined as the conducting peers must satisfy the demand of reliability during the retrieving processing, while the response time must be shortest and the systems' resources must be efficiently consumed. (â…´) Based on the key techniques described above, we have implemented PeerIS: a prototype of P2P-based information retrieval system. In a word, this thesis has detailed the design, key techniques and experimentalresults of a P2P-based information retrieval system, which have been realized inPeerIS. All these contributions are achieved through a comprehensive study on therelated theories, existing technol...
Keywords/Search Tags:peer-to-peer computing, information retrieval, resource location, query routing, queried result ranking, retriving optimization, fuzzy peer reliability analysis
PDF Full Text Request
Related items