Research On Data Sharing Scheme In P2P Based Live Streaming System

Compared with the traditional Client/Server model, through the use of clients'spare bandwidth, the Live Streaming System based on Peer-to-Peer (P2P) network reduces the load of server efficiently, improves the system's scalability, and also raises the system's overall service ability. But with the features as large amount of nodes, high dynamic, high heterogeneity, high asynchrony, it increases the difficulty of system design.An efficient and stable data sharing scheme is the core of P2P based Live Streaming System. Data sharing is through the P2P network topology organizations logic, to achieve efficient, rapid, stable data distribution and transmission between nodes, it include the buffer management, topology maintenance, and data requests scheduling. Existing data sharing scheme focused on the system topology design, few about buffer management and scheduling strategy only solve distribution of media data with a constant bit rate. When face to media data with variable bit rate or large bit rate, it has problems like big spending of data retransmission, inaccurate of scheduling algorithm, poor QoS (Quality of Service) of customers.In order to solve this problem, through the improvement of buffer management and scheduling strategy, a novel and practical algorithm is proposed: A two-tier organizational structure of the buffer data, both segments and packets can be shared, improves the efficiency of data sharing, and reduces the cost of retransmission; Incremental buffer mapping table with real-time index, makes clear the amount of data to be scheduled before the scheduling, improves the accuracy of data scheduling, and reduces the control cost; Integrated scheduling strategy by partition, assures the QoS of client peer, and the overall system efficiency of the data distribution; Bandwidth adaptive scheduling algorithm, dynamically adjusts the maximum capacity and the remaining capacity of peers, improves the efficiency of data scheduling, and avoids too concentrated request on some nodes, to some extent, achieves load balancing; Congestion control algorithm of processing requests, raises the efficiency of overall data distribution, and reduces the waste of network bandwidth brought by the overtime data transmission.The improved data sharing scheme has been applied in AnySee system (4th Edition). Testing showed that the new strategy improves the accuracy of scheduling, reduces the proportion of scheduling failures. Especially in the situation with variable bit rate or large bit rate, it raises the efficiency of overall data sharing, and guarantees the QoS of customers.