Robust Dynamic Bandwidth Allocation Method for Virtual Networks

Min Zhang,Chunming Wu,Qiang Yang,Ming Jiang

DOI: https://doi.org/10.1109/icc.2012.6363896

2012-01-01

Abstract:Multiple virtual networks sharing an underlying substrate network is considered a promising tool to diversify and reshape the future inter-networking paradigm. As a simple and straightforward approach, the static bandwidth allocation in virtual networks (VNs) can often be inefficient in practice, and the adaptive allocation scheme with a small time-scale may lead to the transient network and service instability. Due to the fact that the traffic patterns vary over time, the challenge still remains to meet the expected resources allocation whilst promote the network scalability and robustness. In this paper, based on the robust optimization theory we present a robust dynamic approach which periodically identifies bandwidth allocation to VNs to work reasonable well for a range of traffic patterns over a period of time, rather than certain traffic pattern instance. This problem is formulated as a robust optimization problem using path-flow model aiming to compute the minimum-cost bandwidth allocation. Through the primal decomposition, we present a distributed algorithm which consists of two components running in individual VNs and the substrate network respectively. The numerical result obtained from simulation experiments demonstrates the strength and the effectiveness of the proposed algorithm in terms of convergence and acceptance ratio.