Research On P2P Streaming Key Technology And Incentive Mechanisms

With the high popularity of broadband Internet and the rapid development of multimedia technology, it has become one of the killer applications for providing media streaming to a large-scale number of audiences like IPTV in current Internet and, especially, in next generation Internet. Large-scale network media streaming is faced with two critical requirements:scalability and quality of service (QoS). As the client/server (C/S) model is plagued with the scalability problem, IP multicast is restrained by the deployment issues, and content network distribution (CDN) is constrained by enormous investment and maintenance costs, peer-to-peer (P2P) streaming technology has been regarded as the most promising solution for large-scale media streaming over the Internet.P2P streaming technology has made significant progress over the past decade, and a great variety of P2P streaming systems like PPlive, PPStream, UUSee, and SopCast have gained successful and widespread deployment on the Internet. In 2005, BBC carried out the living TV experiments with iPlayer using P2P streaming technology. In 2009, CNTV also provided the large-scale TV streaming service with CBox based on P2P streaming technology. The measurement results of PPLive showed that there were more than 200,000 users simultaneously watching the Spring Festival Gala in early 2006, and that the total aggregate bandwidth amounted to 100Gbps while the total aggregate bandwidth of Akamai network reached only to 300Gbps in late 2006. In addition, the statistical result of a China ISP's backbone traffic showed that PPLive accounted for 10% of the whole traffic volume that already surpasses BitTorrent traffic share (8%), and that demonstrates the tremendous success of P2P streaming technology. Hence, there is no doubt that P2P streaming technology provides a new opportunity to overcome the problems and challenges in large-scale network media streaming. With the alluring application prospect and marketing potential that P2P streaming technology shows, its research is becoming more and more important in both economic and social sense.Recently, P2P streaming technology has attracted considerable research interest from the academia and industrial communities. The researchers have carried out a comprehensive and in-depth study of P2P streaming technology, in aspects of system modeling and theoretical analysis, overlay network construction and data scheduling mechanisms, and network performance measurement, and made a number of important and creative findings. While current research has layed a solid theoretical foundation for P2P streaming technology, various measurement results have identified that the quality of service of existing P2P streaming systems is still not so high enough that may even hinder its further growth. The measurement results include:(1) the video quality is poor in that it only supports 450-700Kbps streaming rate and, even worse, it frequently comes across the buffering and rechoking problems; (2) the playback startup delay is as high as 20 seconds to 1 minute;and (3) it is disadvantageous over QoS-guaranted IP multicast based IPTV technology in metrics of playback startup delay, video quality and robustness.It is obvious that P2P streaming technology is still confronted with various open issues and challenges including:(1) overlay network problem. The measurement study of Joost shows that the peers only contribute 1/3 of the demanding bandwidth, while the remaining 2/3 bandwidth is contributed by the streaming servers. Thus, pure P2P streaming technology is insufficient to satisfy the requirements like the quality of service and network management. It is required to design a hybrid structure based P2P streaming overlay network; (2) network heterogeneity problem. Peers in P2P networks are with different bandwidths and thus are apparently heterogeneous; therefore, how to cope with the effect of network heterogeneity is a key issue in P2P streaming systems. Current solution is based on new video coding techniques. As most existing P2P streaming systems are based on mesh structure, it is needed to further study the problem of combing mesh-based P2P streaming systems with new video coding techniques; (3) free riding problem. Peers in P2P systems are generally autonomous, rational and even selfish, thus inducing the free riding problem (i.e., a peer consumes other peers' resources without contribution). Free riding problem is always one of the serious challenges confronted with P2P technology. In contrast to P2P file sharing application, P2P streaming systems are more seriously affected by the free riding problem as they have stringent requirements for continuity and timelineness. Thus, it is essential to further study the effect of free riding problem on P2P streaming systems, and to design appropriate incentive mechanisms to alleviate the impact of free riding problem and improve the quality of service.Therefore, P2P streaming technology still needs to be further studied to cope with the challenges including overlay network construction, network heterogeneity, and fairness problem. In this paper, we are focused on the detailed studies on P2P streaming key technologies and incentive mechanisms, including overlay network construction based on hybrid structure, P2P streaming technology based on scalable video coding, modeling and analysis of free riding problem, and trust and incentive mechanisms. The main contributions of the paper are briefly summarized as follows:(1) We present a P2P streaming overlay network construction method, named HybridStream, based on hybrid structure, which systematically leverages the advantages of P2P, CND and IP multicast. The combination of P2P and CDN can significantly increase system bandwidth and thus improve the peer's quality of service, and the combination of P2P and IP multicast can improve the data transmission efficiency in the dense user mode and lower the backbone traffic volume. We also propose a CDN surrogate election algorithm to reduce the deployment cost of CDN surrogates. The computational analysis of performance improvement and control overhead shows the efficiency and feasibility of the proposed method.(2) We present a P2P streaming system design based on the scalable video coding. We theoretically analyze that higher control overhead and more redundant data packets are induced when the peers are constrained by downloading bandwidths. To solve this problem, we incorporate scalable video coding into mesh-based P2P streaming technology. We describe P2P overlay network construction method and data scheduling mechanism, and also propose an adaptive video layer selection algorithm. The simulation results indicate that using SVC can efficiently reduce the control overhead and redundant data packets, thereby significantly improving system efficiency and video playback quality when the peers are constrained by downloading bandwidths.(3) We present a simple probability model to study the free riding problem in P2P streaming systems. We theoretically analyze the impact of free riding on system performance including efficiency, playback continuity and transmission latency, and we also discuss the relationship between the impact of free riding with specific system parameters like server bandwidth, peer degree and buffer length. Through some numerical results, we provide a quantitative understanding of the impact of free riding and conclude that it is essential to introduce incentive mechanism in P2P streaming systems.(4) We propose a content differentiation service based incentive mechanism, wherein peer's reputation is computed with two factors:its contribution level and the popularity of the program, then a differentiated program list is returned to the peer in accordance with its reputation value. Hence, the free-riding problem can be restrained by the implementation of content differentiation service. The simulation results demonstrate that the proposed incentive mechanism can efficiently improve the playback quality in P2P streaming systems.