Recently, Peer-to-Peer network technology has been a hotspot in computer network technologies. P2P network is a distribute system, every node can act as both client and server, they aren't controlled uniformly, and they make up a whole system. Now P2P technology has widely applied to file sharing, cooperation works, distribution computing, and file sharing based in unstructured P2P network is the best popular application. It's a basic technical problem that how to find the node which has special data. The different system adopt different searching algorithm, searching algorithm is related to the efficiency, network traffic and scalability of Peer-to-Peer system.It is because the model of the unstructured system was designed simply; the node's leaving and in the network was less influence, and also supported the popular keywords searching method that unstructured P2P networks were applied abroad. Unstructured P2P network adopted the traditional Flooding searching algorithm. There existed a great number of redundant messages and thus it would consume a lot of bandwidth and induce delay. Based on the traditional Flooding algorithm, researchers put forward many new searching methods, the Dynamic Searching algorithm make use of the history behavior information of the node to calculate the TTL value of the query packets for the next interaction searching. It could reduce the network traffic and save network spending. But when it was applied to search unpopular file, it could increase the total latency in the searching process. For P2P technology's extensive applications, we haven studyed the characteristics of the Dynamic Searching algorithm, then we designed an improved algorithm of the original algorithm to reduce the latency and minimize the payload of the network.This paper first introduced the basic general knowledge of P2P network technology. Then it studyed the usual searching methods based on the unstructured P2P network, also analyzed all kinds of searching methods in unstructured P2P system which recently apply extensively. We pointed out the disadvantages and advantages of some existing searching algorithms. After laboring the Dynamic Searching algorithm, we use the knowledge of the probability and statistics also the characters of the network's topology and node to purify our former Dynamic Searching algorithm. The improving works based on three fields are: Firstly, in the beginning probing process, we decreased the TTL value of the querying packet to control the transmitting of redundant messages, so it could reduce the network traffic; Then, we took an interval estimating method of the probability and statistic to estimate the popularity of the file more safely and correctly, it made the TTL value of the querying packet more reasonable. so it could reduce the latency and minimize the network traffic and save payload; Lastly, in the iteration searching phase the source node only choose one neighbor node not three to transmit the querying packet, it could make the searching scale more stable and make the searching scope more moderate, so it improved the searching performance. They were the emphases of this paper .At last this paper modeled the former Dynamic Searching algorithm and improved algorithm in the simulators. Simulation results show that this algorithm was better in enhancing the network expansibility, saving spending and reducing latency. |