QOS path provisioning in inter-domain network with game theory

Abstract

Providing a guarantee for end-to-end QoS in inter-domain networks is a challenging problem due to the lack of cooperation among ISPs. In this dissertation, a new framework based on a network environment with non-cooperative ISPs has been proposed for apportioning of ISP's responsibility in an end-to-end QoS request. The newly proposed QoS provisioning framework with Path-Classification scheme under Nash equilibrium (PC-Nash) is obtained by classifying the paths according to the QoS-level and optimally selecting the QoS-level apportioning at the Nash equilibrium. A loss network model has been formulated to promptly calculate the call acceptance probabilities as well as the resultant expected utility value of the game solution at Nash equilibrium indicating the optimal QoS-level apportionment of ISPs. PC-Nash has been analytically evaluated and compared with three conventional policies (most-effort, least-effort and equal-distribution). The results show the conformity of call acceptance probabilities between mathematical analysis and simulation at 95% confidence interval. Based on the utility functions of practical ISP business models (i.e., peer, retail and wholesale service models), the experiments demonstrate that PC-Nash outperforms the conventional policies. However, most-effort and least-effort policies provide comparable utilities to PC-Nash with respect to peer and retail/wholesale service models, respectively, for the network with the same path quality. Further, the utilities of all conventional policies are significantly less than the utility of PC-Nash for the networks with different path qualities. By these evidences, the framework of PC-Nash is expected to be most useful in QoS provisioning trials of practical inter-domain networks in the future.