Research On Peer-to-Peer Computing For Overlay Network-oriented Typical Applications

Overlay network theory and technology have not only been a hotspot in theory research field, but also caught the intensive attention of business circles. They have been applied in many fields, and have achieved substantial results mainly in the fields of distributed computing, file sharing, collaborative computing and so on. The further development of the application of the overlay network could greatly enhance the social productivity, improve the social living quality and therefore make bigger social progress.Alone with the deepening of the research and the extension of the application, many new problems and challenges, which have never arisen in IP network, have emerged in the field of overlay network, causing higher requirements of the overlay network. Experts now begin to pay attention to the new application and theory research in this particular field. This article is supposed to do the research on the new development and application of the unstructured overlay network topology model with the reference of three hotspot applicationsThe key issues this paper want to discuss include: any-source overlay multicast in the application layer communication group that has the capability constraining function; improvement of the performance of the overlay search technology of shared files; balanced distribution of search overhead and reasonable balance between search overhead and maintenance overhead in web text information search, and etc. The research details include:(1) Discuss the any-source overlay multicast technique and related technology on application layer based on the overlay network. This paper put forward an any-source overlay multicast service solution without the need to establish a visible multicast tree. This solution is based on the random overlay network and non-DHT overlay network topology model, having the node capability-constrained function as well as low maintenance cost of multicast tree and dynamic member management ability. In this paper two distributed multicast algorithms are designed. We have made theoretic analysis on balanced performance of search hops and search overhead, obtained results that the expected hops which will transfer any-source multicast information to all nodes is O(log_c n), the search overhead is n + O(log_c n),in which c is the average node capability, n is the number of the nodes in the multicast group. Simulated experiment shows that new any-source multicast algorithms have better integrative performance(2) Discuss the performance problem of shared file search tool in overlay network and its relevant solution.This paper paid most attention on the research of the improvement of search efficiency within the unstructured overlay network which has no centralized index structure. This paper found out four problems from KaZaA and Gnutella that affect the performance, and furthermore gave a solution for every problem. This paper brings forward a search algorithm based on the ticket in order to decrease excessive search overhead; the proposed random sampling solution reduces the search latency caused by timeout between consecutive TTL-constrained floodings. And the random anchor solution is used to reduce the repeated visit to a same node. Besides, this paper, based on the random overlay network and non-strict loop, puts forward an overlay search service dealing with non-popular shared file query requests. Related theory analysis and simulated experiment in this paper showed these solutions could preferably improve the file search performance.(3) Discuss the way to efficiently combine the advantages of unstructured overlay network and structured overlay network for the purpose of designing semantic network topology that could meet the web text information query requirement. This paper combines the characteristics of web text information structure and semantic similarity together to build a text information query-oriented semantic network topology. By referring to the IP routing technique, this paper puts forward a two-layer query routing architecture. This architecture consists of core routing layer and edge routing layer. Routing information is entirely distributed on the nodes over the two layers, and there is no central infrastructure. It could be used in large-scale distributed system to decide which nodes should accept query, and therefore realize the query routing function in semantic network. Meanwhile, this paper advanced a concept of preference community, which could effectively organize the web text data source to simplify the query routing information, ease the organization and maintenance and improve the query routing efficiency. Analysis on the time complexity of the algorithm shows it hasgood scalability. The result of the web text information query is obtained with high efficiency using the distributed data filter algorithm--TOP-N algorithm . In addition, relevant technology could show its advantage in the field of overhead balance, integrality of query results and etc.