| Summary: | 碩士 === 國立臺灣海洋大學 === 資訊工程學系 === 101 === In recent years, the network technology is growing rapidly that result in information media resources are increasing quickly. The problem of unbalanced loading on client-server model makes the Pee-to-Peer getting more attractive. P2P network can not only provide file-sharing, but also be used in instant messaging like Yahoo Message and Skype.
The search algorithms on P2P networks can be classified into two categories, blind search and knowledge-based search. This study focuses on blind search, in which Flooding and Random Walk(RW) are two popular ones. Flooding operates by sending query messages to all neighbors and generates a large amount of query messages. On the other hand, RW only forwards a query message to a randomly chosen neighbor at each step, and the response time and success rate are bad. The Dynamic Search(DS) algorithm was proposed to overcome the problem above, which is a generalization of Flooding and RW. It uses Flooding for short range search and RW for long range search. Nevertheless, the search efficiency of DS has faced some bottleneck.
This study propose some idea to deal with search problem. First, to reduce redundant query messages, Flooding is replaced by Probabilistic Search. Second, to enhance the success rate, the RW is generalized to a layer-dependant k-walker, in which each layer has a specific k value. Third, to dynamic adjust the amount of query messages, the k-value on each layer is determined by previous layer, i.e. Balance Search.
This study simulates the combination of these three ideas. First, we combine Probabilistic Search and Sequence Search such that the Probabilistic Search is used in short range, and the Sequence Search is used in long range. Second, we combine Probabilistic Search and Balance Search. The search efficiency and success rate are improved when search efficiency does not consider response time. At last this study uses Iterative Flooding Search to improve the success rate of DS. According to the experimental results, Iterative Flooding Search’s success rate and search efficiency is the best.
|
|---|
