Hybrid Path Selection and Overall Optimization for Traffic Engineering

Abstract

Due to the innovation and development of network technologies, many services and applications have emerged with diversified requirements. Therefore, network operators need to dynamically select paths with different Service Level Agreements (SLAs) to satisfy the services and applications. Traditional path selection algorithms mainly deal with single routing criterion, which is insufficient in this context. In addition, network operators need to optimize the traffic distribution of all flows with limited network resources. In this paper, we propose a method of hybrid path selection for individual flows and overall optimization for all the flows. In our method, we use different algorithms to find paths for two types of traffic, SLA-required traffic and background traffic. The paths for SLA-required traffic are selected based on multi criteria such as delay, cost, and availability, and that for background traffic are based on K shortest paths (KSP) regarding one criterion. Moreover, we consider optimization of the overall objective and determine the bandwidth allocation on the selected paths by linear programming and a greedy algorithm. Specifically, we consider a typical traffic engineering scenario: minimizing the maximal link utilization. The results show that our method is better than Equal Cost Multi-Path with KSP in terms of the amount of satisfactory SLA-required flows and minimization of maximal link utilization.