Research On TeXCP Under Uncertain Traffic Demand

TeXCP is a load balancing algorithm for best effort traffics in MPLS network which presented in Sigcomm 05. It has stability of control theory and convergence to the equilibrium point quickly compared with other algorithm. But TeXCP only balance the traffic demand in the pre-established path set. If the traffic demand is out of prediction their will be a bad performance. COPE(Common-case Optimization with Penalty Envelope) guarantee worst-case performance which was presented in Sigcomm 06. But COPE is an off-line algorithm, can't converge quickly as on line TE algorithm.In this paper, we import the Penalty Envelop in the COPE into TeXCP and present a unpredictable traffic demand TE algorithm– PETeXCP. The basic starting point of PETeXCP is that it can be automatically adjusted to reach multipath load equilibrium by using TeXCP under predictable traffic demand and adjust path set under unpredictable traffic demand so TeXCP can work on new path. PETeXCP is a dynamic load balance algorithm which guarantees worst-case performance.PETeXCP made some change and extension on TeXCP: 1, an increase of the management of the path set, change path set according to real time network link status; 2, the expansion of the TeXCP agent probe function, the agent can not only probe the active path but also the backup path and the inactive path; 3, modification of the TeXCP algorithm, to enable it to choose new path dynamically and make load balance in the new active paths.This paper also implemented PETeXCP in NS-2 and MNS2.0 simulation environment. The simulation results show that the new algorithm effectively reduce the maximum link utilization and reduce the congestion of the network when the unpredictable traffic demand turns up to ensure acceptable performance when the network experiences unpredictable changes.