Research On Several Key Performance Optimization Technologies For P2P System

P2P (Peer-to-Peer) system provides an ideal framework for fully utilizing the network edge computing resources. It has been widely used in many fields including content sharing, information retrieval, instant messaging, collaborative computing, and has been regarded as one of the most important technologies affecting the future of the Internet. The autonomy and openness of P2P networks makes it more difficult to carry out effective security controls, so disignning tools to simplify development and deployment of P2P systems can help to improve monitoring of P2P networks and promote the further development of P2P systems. P2P traffic has occupied the main part of the Internet and a large number of P2P traffic impacts the bearer network very much which affects the quality of services of other Internet applications seriously. Optimization of P2P traffic will help improve the bearer network bandwidth utilization. The speed of P2P systems distributing data is one of the most important indicators of performance, and improving the speed and fairness of distribution can service P2P customers better. In order to improve the performance of P2P application, such concurrent technologies as multi-process and multi-thread are needed. The advanced concurrency control method will help protect the concurrent performance of P2P systems. For these above purposes, several key performance optimization technologies for P2P systems have been studed in this dissertation, from macro framework to the implementation details. The contributions are as follows:First, the P2P Oriented Security Service Platform (POSSP) is designed to simplify the development and deployment of the secure P2P systems. The platform consists of three partition including network control center, tracker overlay network and the shared server cluster, the overlay networks of control center and tracker are organized on distributed hash table technology, and the functions of P2P basic services and advanced security are provided in web services. The new P2P applications can be deployed on the platform freely and the fast non-repudiation of the symmetric encryption communication are supported in POSSP. The qualitative analysis shows that POSSP can simplify the development of the safe P2P systems and prevent malicious nodes, resist Dos attack, realize supervision effectively.Second, the genetic algorithm based cooperative relay strategy is proposed to optimize resource utilization, improve the global distribution performance, and reduce the network bandwidth consumption of P2P. Firstly the data distribution mechanism of cooperatively relaying between different networks and fetching the data as near as possible reduces the P2P traffic effectively. Then, the constrained multi-objective model of cooperative relay is derived and the optimization algorithms' requirements are also proposed. The single objective genetic algorithm (SOGA) and the multi-objective non-dominated genetic algorithm (NSBGA) are designed. With respect to the model of cooperative relay, NSBGA customizes the initialize policy and genetic operators, which has better performance than the original multi-objective constrained optimization algorithm. The performances are verified and analyzed in the simulation.Third, a P2P traffic optimization schema of Status and Locality Aware Operator Cooperative (SLAOC) is proposed to speedup data distribution and reduce P2P transit traffic. The sensor firewall is designed in SLAOC, which can help to realize the P2P registration and classification, so it helps to implete operation and encourage P2P users to participate SLAOC. The sensor firewall based distance calculation method is proposed, using congestion conversion algorithm the communication cost evaluation integerates the location and network connection status, and the trackers, which are constructed to an overlay network based on the distributed hash table, provides standard web services for the P2P communication cost query. The tracker node selection, peer node selection and the uploading and downloading control algorithms of Bittorrent are improved, and the improved algorithms can be used in other traffic optimization architectures. The simulation results show SLAOC can speed up peers' downloading and reduce inter-domain traffic, and respond to network congestion automatically. A win-win case is achieved in operators and P2P users.Finally, starting from the examples, the Concurrent Computation Redundancy Problem (CCRP) is defined systematicly. The traditional concurrent control methods can not directly avoid duplicate computing which may decrease the concurrent performance. The related performance index and judging methods are given, and the general concurrent redundancy control mechanism is designed based on Producer/Consumer model. The single condition and cross condition redundancy control algorithms have different application scopes and can be used as fixed patterns to solve CCRP. Relative property proofs and simulations are given based on Petri net. The concurrent program experiments show the concurrent redundancy control is necessary and efficient. Two control algorithms are compared in the experiments. The research is discussed on the general concurrent program and so the conclusions have universal reference value for the data race detection and concurrent programming.