Mapping bandwidth to quality of service: an importance sampling based traffic engineering approach

This paper proposes a new traffic engineering approach: importance sampling based traffic engineering (ISTE). ISTE can map the bandwidth of a traffic flow to the quality of service (QoS) it can receive within a network. The proposed ISTE approach does not require extensive knowledge of the network internal details, thus making it applicable to most large and complex networks. It can carry out the end-to-end QoS analysis of a network or carry out the performance analysis of a single network node. Even if there are multiple congested nodes in the network, the ISTE approach remains effective. This paper shows that the ISTE approach, under self-similar (C. Huang, Ph.D. thesis, Carleton University, Canada, 1997; V. Paxson and S. Floyd. IEEE/ACM Trans. on Networking, vol. 3, no. 3, pp. 226-244, 1995) traffic model, is capable of calculating the changes in the network QoS (e.g. probability of buffer overflow) with respect to the changes in the bandwidths of the ingress network traffic flows. In the scenarios where several ingress traffic flows influence the QoS of the network, a more specialized technique called ISTE alternating twisting (ISTE-AT) is proposed. ISTE-AT makes the proposed ISTE approach even more powerful.