On Some Key Techniques Based On Semantics And Node Storage Capacity In Unstructured P2P Networks

In recent years, network technology developed rapidly. Network becomes an increasingly important part of our daily life. The massive resources in the network show exponential growth with the rapid expansion of the network. Therefore, efficient resource search has become a pressing problem. P2P resource search is one of the core problems and it is also a hot point in the research area of P2P networks.For the traditional search methods, due to a wide range and a huge amount of resources in P2P networks, the efficiency of resources search is very low, and the cost of the search is giant. This has become one of the important factors of limiting the development of P2P networks. Reasonable P2P networks topology can effectively reduce the search costs and improve search efficiency. At the same time, efficient search strategy can further improve search efficiency and reduce costs.By analyzing the lack of traditional topology construction methods, search strategy, and adaptive file replication method of hot file, and based on the analysis of current status and trend of resource search, in this paper, we have made some depth study about search performance of P2P network topology construction, search strategies, and adaptive replication of resources of hot file, and proposed some solutions. The contributions of this dissertation include:1. Proposed a P2P network topology construction method based on semantic and nodes'storage capacity. P2P network topology is the basic guarantee for the efficient search and location of resources. In literature, little attention has been paid on the semantic relationship between nodes and node storage capacity in the P2P network topology construction. Accordingly, such network topology leads to the inefficient resources search. This paper presents a semantic correlation calculated algorithm which jointly considers both of the storage capacity of nodes and semantic key words in nodes. Therefore, this method can achieve a more optimal P2P topology. Experiments show that our algorithm makes the P2P network build a more reasonable topology. It has higher resources search efficiency and lower search costs.2. Proposed a search method BF-SKIP (Biased walk, Flooding and Search with K-Iteration Preference) which based on semantic and nodes'storage capacity. P2P network search strategy is another basic guarantee for the efficient search and location of resources. Existing studies neglect the semantic relationship between nodes or they simply adopt the random walk to search resources. In this paper, under the circumstances of considering the network topology with semantics and node storage capacity, we proposed a three-stage search strategy to achieve efficient resources search, and low costs. First, locate a related semantic group based on semantic relationship; then search within this group (through semantic links) only one hop in according to the most relevant principles; finally, SKIP mechanism will continue this search process. The simulation results show that the BF-SKIP scheme fully takes semantic into account and the search efficiency of resources have been further improved and search costs have been further reduced as well.3. Proposed an adaptive file replication method based on semantic dynamic community. Hot file replication method is another important factor to improve the search efficiency and reduce the search costs. In previous studies, little effort has been devoted to prediction function and the semantic correlation between nodes in eliminating access hot spots strategy. Accordingly, it cannot effectively put the copy on the demand area, and there is a great blindness for placing a copy. That causes a lot of search costs and low search efficiency. This paper introduces the semantic correlation of nodes and the prediction function. Prediction function can effectively predict hot resources in advance, and copy the hot resources in advance. Then it can place the copy in a more appropriate location by semantic correlation of nodes. Experiments show that it can dramatically improve the replica hit rate, get lower search costs and reduce the overhead of file replication. That means it reaches relatively high query efficiency with comparatively low replication overhead.