| Routing hotspot, a.k.a. flash crowding, is one of fundamental issues in P2P streaming distribution network. A routing hotspot is typically created by an unanticipated new event that triggers an unanticipated surge of users that request streaming service from some particular peers, temporarily overwhelming the peer's delivery capabilities and overloading network connections. However, for P2P streaming distribution network, it is difficult to resolve the routing hotspot problem only depended on the current routing mechanisms. Thus, it is necessary to design a series of schemes to resolve the routing hotspot problem in P2P streaming distribution network and distribute media streaming contents with high quality.In this thesis, we resolve the routing hotspot problem in P2P streaming distribution network from the following aspects: (1) how to avoid the routing hotspot prior to a hotspot event; (2) how to discover the routing hotspot location fastly; (3) how to recover the routing hotspots efficently. In summary, the major contributions of this thesis are as follows:(1). We consider this hotspot avoiding problem from a game theoretical standpoint. The problem is posed as a noncooperative game, for which the Nash equilibrium was investigated. We propse an analysis model for the problem. Based on the model, we propose an incentive-compatible pricing mechanism, which can driver the network to the Nash equilibrium. We also propose an adaptive algorithm for distributed computation of the incentive compatible pricing mechanism. The theoretic and simulation results show that the mechanism and algorithm can avoid the routing hotspot efficiently.(2). We consider how to discover the routing hotspot location. First, we propose a graph-based system model, which takes into account the dependencies among multiple different peers and is suitable for hotspot localization in large-scale real-time P2P networks. Second, based on this model, we prove that the problem is NP-Complete and thus design a heuristic algorithm (MMDHD algorithm) for finding a near-optimal solution to hotspot location. The theoretic analysis shows that our algorithm is a near-optimal solution from the two aspects: computational complexity and the accuracy of hotspot location. Finally, the extensive simulations are also conducted to verify that the MMDHD algorithm can locate routing hotspots quickly and efficiently. (3). We address the problem of efficient hotspot recovery in P2P streaming distribution networks. We propose layered routing hotspot recovery (LRHR) mechanism and also present two variants of LRHR: GNP-LRHR and RRNS-LRHR. GNP-LRHR construct Delaunay Triangulation (DT) meshes among the hosts by the Bowyer-Watson algorithm and then find the optimal hotspot recovery scheme using the property of DT. RRNS-LRHR selects the recovery neighbour randomaly. We present the analysis on the complexity and hotspot recovery delay on LRHR and the results show that LRHR is an effective hotspot recovery mechanism. The simulated results show that LRHR is effective to recovery the routing hotspot.(4). We address a typical application instance of P2P streaming distribution network: the video conferencing based on P2P-SIP. We propose a dynamic scalable framework of P2P-SIP video conference. Based on the framework, we implement a prototype of P2P-SIP video conference: SOPVC. The SOPVC adopts the single-speaker conferencing mode. Based on this mode, we simply experiment these proposed mechanisms. The experiment results show that these mechanisms are efficient to solving the routing hotspot problem in the P2P streaming distribution network.
|
PDF Full Text Request