Unequal Block P2p On-demand Technology And Its Scheduling Problem With

With the widespread deploy of Internet, the population of Internet users become greater and the demand for Video-on-Demand services keep growing. Video-on-Demand is a service demand for high-bandwidth and low latency, thus the traditional Client/Server architecture system can not serve numerous users at the same time because the server resources is the system performance bottleneck. However, in P2P system each user serves the other users when he uses other users'service. P2P technology makes use of resources at the edge of network, which shares the pressure on server. The researchers introduced P2P technology into Video-on-Demand system; this makes P2P Video-on-Demand system emerged. P2P technology brings new development for Video-on-Demand, which makes possible to build large-scale P2P Video-on-Demand system, so it is worth to make deep research on P2P Video-on-Demand system.High real-time and strict order is required in Video-on-Demand system, additionally; the user may skip to another point, which brings additional data transfer for the stabilized server. Based on analysis to traditional P2P Video-on-Demand technology block algorithm, an uneven block algorithm was proposed in this thesis. To reduce server load to, an optimize solution of schedule was given and the parameters of optimize algorithm was analysised. The details are as follows:(1) The features of anchor technology was studied, an uneven block algorithm was proposed against current even block system and an example was provided (without consider of downloaded data from the server). The prove of using this algorithm can reduce the delaytime when one skip to another point under the same condition of segment size and number of block in even block system was given, meanwhile by discussing the adaptation of this algorithm, the conditions of smooth play under other environmental was concluded.(2) The features of improve Particle Swarm Optimization-Uniform search Particle Swarm Optimization (UPSO) was analysised, the UPSO algorithm formula was transformed into a non-recurrence location update formula, and the UPSO's convergence region to learning coefficient c and inertia coefficient w was concluded by deducing the solution convergence conditions. To confirm the conclusion, simulation experiments were provided on the selected location of the region of convergency by6Benchmark functions. Experimental results show that UPSO converge when the learning coefficient and inertial coefficient in the convergence region and diverge outside convergence region.(3) By analysis the scheduling sequence of uneven block system, a conclusion that scheduling sequence influence burden on server was obtained, an greedy algorithm based on the finish time after download was put forward to calculate particle fitness. A bijective between the download order and particle's position was found to apply UPSO in scheduling problem. As experimental results shows:the application of UPSO in scheduling reduces the load on server. In a comparative experiment, with the same segment size and number of block in even block system, the server pressures is greater.