Dht-based P2p Resource Locating Model Study

As an important basic service in peer-to-peer applications, on peer-to-peer resource location is always the hot topics. Resource location technique has very important influence on the scalability and robustness of peer-to-peer systems. Compared with the traditional Client/Server system, peer-to-peer system exhibits some special characteristics, such as large-scale, high dynamics, heterogeneity, which is a big challenge to peer-to-peer resource location. Based on the difference of topology, peer-to-peer resource location can be classified to Unstructured and the Structured resource location.Structured resource location techniques are mainly based on the distribute hash table, which can route to the correct resource in o(log n) hops. Unstructured resource location techniques are mainly based on the flooding search, which has good fault-tolerance in the network churn without maintenance cost and good effectiveness and quality in searching the hot resource.To solve the problem above, this thesis proposes HRLM, a hybrid resource locating model. HRLM utilizes the heterogeneity in different peer and divides the peers into two types—ultrapeer and leaf peer. The ultrapeer takes charge of locating the resource and maintaining the network. And the leaf peer who wants to participate in the network should connect to the ultrapeer. In order to keep low overhead, HRLM chooses correct resource location from the popularity of the resource. If the resource is hot, HRLM searches it with flooding. Otherwise, HRLM searches it with Kademlia. In order to improve the performance, HRLM also use assistant locating which is local index and assistant locating. Local index caches index information about which neighbor node have which files, which can reduce the amount of the search message and keep low search latency. Bloom Filter compresses the information which is received from the neighbor to keep low overhead. The experiments show techniques can improve the resource locating performance of P2P system efficient and can be applied to large-scale P2P systems.