The Research On P2P Media Delivery Technology

With the rapid development of the Internet, the media streaming technology has achieved great progress. On the other hand, with the emergence of peer-to-peer (P2P) technology, network bandwidth and the capacity of user end-systems which were ignored in the past become valued computing resources now. How to utilize end-systems resources to build a P2P media delivery system with large-scale, high scalability, reliability and performance in a dynamic peer-to-peer network environment is a hot research topic in recent years.Based on a systematical summary of relevant works on media delivery technology, this dissertation focuses on related technologies of P2P media delivery and gains several achievements on some sub-topics. The major contributions of this dissertation are as below:1. This dissertation proposes a P2P-based large-scale media delivery network, PPMDN. Based on its hybrid P2P media delivery architecture, PPMDN organizes large numbers of end-system nodes into a united overlay network and builds a P2P-based media delivery system with high scalability. Featuring with effective live video delivery on bone networks, distributed topology matching mechanism, server bandwidth distribution algorithm, distributed caching management mechanism, PPMDN provides users a large-scale media delivery service with highly guaranteed QoS.2. This dissertation presents GA-MWPL-DC-ST, an optimization algorithm of overlay multicast network for live video aiming at decreasing the latency of lost synchronization in living video distribution. GA-MWPL-DC-ST algorithm constructs minimum weighted path latency degree-constrained spanning tree (MWPL-DC-ST). Furthermore, heuristic algorithms are adopted in phases of initialization, crossover and mutation during the genetic iteration, and an adaptive parameter according to different evolving generations is used to accelerate the convergence of the algorithm. Simulation results show that the GA-MWPL-DC-ST algorithm can get better optimization solutions than other heuristic algorithms.3. This dissertation presents a distributed topology-aware node rendezvous algorithm, TANRA. In P2P streaming, the overlay network consists of peers may mismatch the underlying physical network, which results in larger media transmission latency, low efficiency and high bandwidth consumption. Accordingly, based on network coordinates and distributed hash table, TANRA, the distrbibuted topology-aware node rendezvous algorithm is proposed. The unique mapping between 2-dimension network coordinates space to multi-layer name space in Chord can keep the proximity between nodes. Furthermore, nearby neighbor nodes can be located quickly through Chord distributed object routing mechanisim. This will provide fast topology matching in the large-scale media delivery service.4. This dissertation presents a server bandwidth allocation algorithm of P2P VoD. Current P2P VoD solutions adopt the best-effort way to schedule server bandwidth. In this way, however, the server bandwidth resource cannot be efficiently utilized and the global VoD serving capacity cannot be quickly scaled. Therefore, after constructing the theoretic model of cooperative media delivery in P2P VoD, this dissertation proposes two algorithms: DP-BDA algorithm based on dynamic programming and CB-BDA based on heuristic climbing. In CB-BDA, the server bandwidth is optimally allocated according to the service capacity and the bandwidth requirement of one video file. The sever bandwidth for video files with high visiting rate and low service capacity can be allocated preferentially. Thus the entire system service capacity can be quickly promoted and user experiences can be improved.5. This dissertation presents a distributed cache management mechanism for P2P VoD. Adopting centralized and hierarchical tree and replacement algorithms based on visiting rate, current cache management proposals cannot efficiently utilize cached video segments in user end-systems and increase system service capacity. In this dissertation, a novel distributed cache management mechanism is proposed. It includes a distributed cache management and replacement algorithm based on Chord and dynamic programming. The distributed object routing mechanism of Chord can not only improve scalability and resilience but also accelerate the search of video segments. Moreover, by searching optimal cached list of video segments, DP-CRA algorithm can ease the VoD server (ESPV) pressure and improve the QoS of both "hot" and "cold" video segments.