Adapt To Changes In The Physical Network Status Into Application Resource Allocation Mechanism And Its Applications

Nowadays, Internet is able to carry voice communication (VOIP), live video and video-on-demand service (IPTV) in large scale throughout the world. They exploit novel resource allocation protocols for content delivery and system control. The protocols are obviously different from that used by traditional unicast applications. We call it per-group application resource allocation mechanism. Traditional applications are usually based on TCP resource allocation mechanism by unicast transmission manner. We call it per-flow unicast resource allocation mechanism.This paper focuses on key technologies of per-group resource allocation mechanism. We extend the study in four aspects including theoretic model, performance evaluation, optimization method and control mechanism. The main contributions are as follows:A. To meet the requirement from both content provider and Internet Service Provider(ISP),we propose metric of "Service Efficiency(SE)". It can describe the relation between provision cost and system capacity. We give the mathematical expression of SE for typical resource allocation models. The SE is determined by some key metrics. We give the optimization principle for the metrics. And this is the basis for optimization protocols designed and implemented in Chapter V and VI;B. According to the network utilization manner of resource allocation model, we classify the system into five types. We conduct a simulation evolution on the typical types of system. The simulation shows that compared to Fully Unconstrained system the Conditional Unstrained and Hybrid system are more scalable and adaptive to ISP's traffic policy. Their service efficiency are better;C. Based on real-time minimum average source-to-end latency topology organization theory, we first propose a contribution-aware overlay optimization method for mesh-based P2P live streaming system. We have designed a suite of protocols to construct and optimize the overlay. Our core idea is to promote nodes with larger contribution closer to the source node so as to keep a compact and efficient mesh-based streaming overlay. Our method is evaluated through extensive simulations. Experimental results show that our optimization method can efficiently exploit the serving capacity of peers so that the performance of the overall streaming delivery is improved. Compared with the non-optimized overlay, the P2P streaming applications using our method could cut down the average latency by 20-50% and enhanced the continuity index up to 120%;D. Propose a united resource allocation architecture called Pharos by collaboration of both ISP and P2P. In the architecture, we establish overlay path metrics system to expose ISP traffic policy and physical link states to P2P. We facilitate P2P system to conduct topology aware overlay routing by peer selection assistance service. We implement the prototype system to integrate biased peer selection and rate control module at peer end. They both respond to network states variation obtained from aforementioned service at ISP end. We verify by simulation that maximum link utilization can be effectively controlled around the objective level. With efficient use of network resource, P2P system also enhances its distribution performance. The completion time could cut down up to 20% .