Research On Resource Discovery For Peer-to-Peer Resource Sharing System

With the explosion of network and enrichment of resources, as a typical instance of distributed resouce sharing system, P2P comes to be the leading pattern of reccent resource sharing. As an essential step of resource sharing, resource discovery dedicates to resolving the mapping problem between distributed resources and users, whose performance directly relates to the effectiveness of resource sharing. Therefore, it is an area that well worths studying.By intergrating the abundant edge resources, P2P enables resource sharing in distributed enviroment. Resource sharing involves three steps:placement, discovery and selection. Resource placement, foundation of the three, focuses on distributing resources onto nodes following the rules of the overlay topology. Resouce discovery intends to locate the desired resources for various applications and is the precondition of resource sharing. Resource selection helps to decide the most appropriate resource. Among these three steps, resource discovery is the key step and closely relates to resource placement. Therefore, the research of resouce sharing should be based on the topology of the overlay.Upon the idea, the dissertation intends to improve the effectiveness of resource discovery of three typical topology of P2P resource sharing systems, which are decentralized, hypersized and constructed, and leads to the research work as follows.We first introduce the P2P file sharing system, including the origins, development and the features. Later then, according to the three listed overlay topologies, we study the problem of resouce discovery, challenge works and unresolved problems.As to the hybird P2P network, we focus on the management of super nodes and resources. For the super nodes, to improve the effectiveness of resource discovery, an AHP (Analytic Hierarchy Process) based super node selecting mechanism is proposed to address the involving multi-factors controlled decision making problem, which is applicable for different scenes. Heterogeous and dynamics of resources complicates the management of resource in P2P network, which lead to high dimension problem and make the range-based resource discovery harder. Addressing such problems, a type-sensitive multi-attribute resouce discovery mechanism is proposed, including a type-sensitive resource management and a Hilbert based resource discovery. In decentralize P2P network, for the inherent blindness of resource discovery, the very option for this problem is to construct an informed search. Traditional informed search tends to maintain the physical layer information such as sharing resources and degree of nodes, and distributes the information evenly in the network, which conflicts with the nonuniform distribution of sharing objects. Addressing the issue, upon the investigation of the behavior of sharing and consuming of resources of nodes, we propose an informed search method named response tracing based resource discovery. By maintaining information of the content and reponser of discovery response, it is testified to be able to cut off the cost and improve the performance of resource discovery.Finally, we analyse the structured P2P network, and draw a conclusion that the main reason for the unsupportiveness of complex resouce discovery in this topology lies in the mapping between resources and nodes. To resolve this problem, based on the theory of linear code, we redefine the mapping and construct a chord-like structed P2P named LiChord. In LiChord, bloom filter is used to maintain the similarity of objects and lay the foundation for approximate match based resource discovery.