IGP traffic engineering : a comparison of computational optimization algorithms

[摘要] ENGLISH ABSTRACT: Traffic Engineering (TE) is intended to be used in next generation IP networks to optimizethe usage of network resources by effecting QoS agreements between the traffic offeredto the network and the available network resources. TE is currently performed by theIP community using three methods including (1) IGP TE using connectionless routingoptimization (2) MPLS TE using connection-oriented routing optimization and (3) HybridTE combining IGP TE with MPLS TE. MPLS has won the battle of the core of the Internetand is making its way into metro, access and even some private networks. However,emerging provider practices are revealing the relevance of using IGP TE in hybrid TEmodels where IGP TE is combined with MPLS TE to optimize IP routing. This is done byeither optimizing IGP routing while setting a few number of MPLS tunnels in the networkor optimizing the management of MPLS tunnels to allow growth for the IGP traffic oroptimizing both IGP and MPLS routing in a hybrid IGP+MPLS setting.The focus of this thesis is on IGP TE using heuristic algorithms borrowed from the computationalintelligence research field. We present four classes of algorithms for MaximumLink Utilization (MLU) minimization. These include Genetic Algorithm (GA), Gene ExpressionProgramming (GEP), Ant Colony Optimization (ACO), and Simulated Annealing(SA). We use these algorithms to compute a set of optimal link weights to achieve IGPTE in different settings where a set of test networks representing Europe, USA, Africa andChina are used. Using NS simulation, we compare the performance of these algorithmson the test networks with various traffic profiles.

[发布机构] Stellenbosch University