Query Mechanism In The Network Study

File sharing applications based on peer-to-peer (P2P) occupy 70% of the overall internet traffic and become one of the most important services in the Internet. Resource query in P2P networks is the core technology in these applications. The peers in P2P networks are composed of millions of common personal computers, which own relative low computing and storage capacity. Therefore, the search algorithms in P2P networks not only support efficient query, but also require low maintenance cost and high fault-tolerance.Unstructured P2P networks usually utilize blind search mechanisms (like flooding). Although the mechanisms can support queries flexibly, they will generate redundant messages in accordance with the exponential growth; Structured P2P networks based on distributed hash table can provide good scalability and high search efficiency, however, it can't support multi-keyword queries directly and its maintenance cost is very high. In this thesis, we mainly study the search mechanisms of P2P networks. The main contributions are as follows.Firstly, a similarity-computing model is proposed for the peers in content delivery networks. The model fully mines the meta-information of files owned by peers, that is, the model calculates the similarity making use of file name, file size, file type according to different weights. Also, the original problem of determining the model parameters is formulated as a constrained optimization problem. Then, the user-interest-based neighbor selection strategy making use of the model is put forward as:trying to make similarity-value-high peers become neighbors, and the eMule application is taken as an instance to show how the model parameters can be determined. Since the similarity-computing model between peers introduces the semantic information of peers, it can be applied to improve the physical topologies of content delivery networks. Finally, the simulation results show that the routing algorithm embedded with the neighbor selection strategy has search performance and large routing improvement, and such proves the validity of similarity-computing model.Secondly, we focus on optimizing the routing performance of structured networks. In view of the advantages and disadvantages of unstructured P2P networks and structured P2P networks in the process of searching resources, dynamic query strategy which adopts in unstructured P2P network is added to hierarchical structured P2P network model ML-Chord. Meanwhile, search range is adjusted in accordance with the popularity of files. In addition, the broadcast algorithm based on Chord is adopted in case that the same node receives duplicate query messages. The experimental results demonstrate that the improved query strategy can not only effectively reduce routing hops and query load, but also shorten time-consuming. Lastly, the conclusion is given and the future research work is pointed out.