Research On Efficiency And Fairness Of Peer-to-Peer File Sharing Systems

The rapid development of P2P technology is allowed to show unique advantages in many areas such as large-scale parallel computing, distributed resource sharing and collaborative work. File sharing system based on P2P technology because of its high availability, scalability, distributed resource aggregation and efficient file transfer efficiency, and therefore won the favor of the majority of Internet end-users and has become the most important means of user to obtain file resources. P2P file-sharing system currently is not only the most important application in the P2P field, but also one of the the main applications in the Internet and has occupied most of the Internet traffic. This dissertation focuses on the research of key technique of P2P file sharing systems, such as resource lookup mechanism, broadcast mechanism, peer selection and aggragate congestion control mechanism. Our researches is beneficial to the improvements of the resource location efficiency and file transfer efficiency of P2P file sharing systems and could grarantee the fairness between P2P file sharing systems and traditional Internet applications. The contents and contributions of this dissertation are as follows:(1) A reinforcement learning based search mechanism (RLBS) for unstructured P2P network is proposed. RLBS build search keyword information table and neighbor information table on the node, and use historical search results to build reinforcement learning model, so as to continuously optimize the follow-up resource discovery efficiency of the system. Simulation results show that RLBS algorithm is able to guarantee query success rate, and not only effectively reduces network traffic, but also reduces query routing hops, thereby improving the system search efficiency.(2) An efficient broadcast mechanism (PTBB) for DHT based P2P file sharing systems is proposed. In PTBB, each node in the system could split the ID space to construct a BPT cooperatively. Through the BPT, nodes could perform broadcast operation rapidly, and will not send any redundant message. In addition, in PTBB we have presented a simple and efficient mechanism to handle system churn that could make sure full coverage of the nodes in the system when the system churn is exist. Simulation results show that, compared with the flooding mechanism of unstructured P2P network, PTBB could achieved(3) An underlying topology-aware peer selection mechanism (UTAPS) for BitTorrent file sharing system is proposed. In UTAPS, Tracker and nodes cooperate with each other to infer the underlying network topology information using network tomography technology. UTAPS could selects peers within small hop counts and low round trip time (RTT) range by utilizing the knowledge of underlying topology. Consequently, the proximities among peers in the P2P overlay network are enhanced. Experiment results show that the UTAPS algorithm can achieve better individual performance in sense of low download time and reduce the traffics which are injected into ISP backbones.(4) An aggregate congestion control mechanism (ACCM) for P2P file sharing system is proposed. ACCM enforces the friendliness of network upon all the connections belonging to an application instead of upon individual connection. By observing the share congestion in access link through application-level measurement technology, ACCM dynamically adjusts the window size of parallel TCP connections, achieving friendliness to tradition Internet applications on the basis of maximize utilization of network bandwidth. The experiments demonstrate that certain fairness and congestion avoidance can be achieved in presence of congestion and the network bandwidth can be effectively utilized in absence of congestion control with ACCM.