多业务量矩阵下te算法性能研究通信与信息系统专业论文.docxVIP

ABSTRACTABSTRACT ABSTRACT ABSTRACT The Interact has gown dramatically over the past 1 0 years．With the fast development of Internet as well as abundance of network traffic and high requir锄ent of network performance，it challenged the network ISP very much．Traffic Engineering is coming．TE is one of the most important network optimization techniques which are mainly dealt wi也optimization for network resources to avoid network congestion． Theoretically,if the TM is known exactly,then optimal routing for it carl be obtained by solving the corresponding multi—commodity flow problem instance．Unfortunately, measuring and predicting traffic demands are illusive problems．Flow measurements are rarely available 011 all links．For a large—scale Internet with bursty demands，it is important for large ISP network operators to find single robust or fixed set of routes to optimize the performance over multiple TMs．Multiple TMs Can be used to characterize the change of traffic or traffic uncertainty in exogenous traffic． This paper concentrates on the topic of traffic engineering based on multiple traffic matrices．In chapter 2，we mainly discuss the problem of performance metric which mclude average case performance,worst case performance and their tradeoff in traffic engineering based on multiple traffic matrices．We propose a now metric。a weighted sum of the average case and the worst case performance-to control血e tradeoff between these two considerations．We prove that optimizing routing using this metric has desirable properties，such as the average case performance being decreasillg，c()ilVeX function to the worst case performance．Then we discuss the metric in our remaining chapters are all about the worst case performance．In chapter 3，the topic is about traffic engineering based on OSPF／IS—IS routing protocol for multiple TMs．We first prove that the traditional TE algorithms for single accurate TM are NP—hard problems not to mention TE algorithms dealing with multiple TMs．We then prov