Design And Implementation Of P2P Stream Real-time Behavior Recognition System In Hi-Speed Network Environment

Because of the resource evanescence nature, the statistic results of the identical of breakpoint and scheduling of the connections could figure out whether a peer has run P2P stream application or not, which is the Breakpoint Scheduling Identical (BSI) algorithm by HUST Network Center 863 project team. Since the flow-based behavior recognition system has problems of scalability, efficiency and coordination in the Hi-speed network environment and the BSI algorithm is sensitive with packet missing, the implementation of real-time P2P stream behavior recognition system in Hi-speed network environment should focus on solving the time and space performance bottlenecks.To improve the efficiency of flow information searching and organization, need to use efficient hash algorithms and optimized hash list structure to manage the flow information records. By extending the range of the hash key can reduce the depth of the collision, using d-left hash algorithm to optimize the hash list structure can make the data distribution more uniform; the complementary of tree and hash list structure could also improve the performance of the hash table. Based on the characteristics of dual-flow, a flow information storage and efficient search method could compress the five-elements to quad, and only need one search can hit the flow.As the traditional heap allocation algorithm has problems of fragmented distribution, extra space, and alignment, and the data backup would really slow. Big Chunk allocation and management technology could reserve data in concentrated memory, which greatly enhance the space performance of flow information storage, and largely improve the time performance of copy and dump of flow records, solving the system scalability issues and coordination problems.Using the efficient data structures and algorithms for flow information organization and searching, and optimized memory allocation and management technology, could implement a P2P streaming real-time behavior recognition system based on BSI algorithm which could run well in Hi-speed network environment. System consists of the packet capture module, the flow information recording module, the flow data backup and transmission module and peer information statistics module. Through the analysis of BSI statistical data to determine threshold, and using the real-set test in actual network, showing that the recognition accuracy of the system is great.