Traffic Engineering In MPLS Network

IP network technologies are evolving in several directions: optical transmissiontechnology develops far beyond, access network technology steadily advances androuters in core networks become bottlenecks. Light weight network technologies thatcan improve network performance, are therefore much needed than ever before. Theresearch work in the thesis is conducted under this background, and the contents arefocused on traf?c engineering in MPLS (MultiProtocol Label Switching) network, in-cluding model establishment, domain selection, routes establishment between nodepairs and traf?c ratio adjustment among several routes. Since MPLS is the basis of traf?c engineering, its performance is also evaluated.In chapter 2,based on DP-Trie data structure and formulas[20],relative MPLS perfor-mance compared to IP is quantitatively analyzed, using classical queueing theory. In chapter 4, a new off-line traf?c engineering model is proposed, in which non-elastic traf?c has high priority. A mathematical model for selecting traf?c engineeringdomain is also presented, along with according algorithms. This is useful to guide thedeployment of off-line traf?c engineering and the design of pratical networks. In chapter 5, the performance of KSP (K Shortest Path) and Link-Disjoint KSP al-gorithms are evaluated, and a new DCDMF (Delay Constrained Dijkstra MaxFlow) al-gorithm is proposed. The new one has better performance, including higher bandwidthusage, fewer required routes and lower average delay, as well as good redundancy inthe routes set. In chapter 6, traf?c mapping technology is studied. Firstly, the difference betweenper-packet and per-?ow mapping is compared, and the primary conclusion is that per-?ow mapping can avoid the TCP multipath problem while sacri?ces fairness between?ows; Secondly, the approximate upper and lower performance bounds of traf?c map-ping algorithms in pure random networks are modeled and analyzed, and the primaryconclusion is that only if the network load is moderate, a well-designed on-line algo-rithm is crucial to improve network performance; Finally, a new Web First AdaptiveTraf?c Engineering algorithm is proposed, and the simulation results turn out that it – II –Abstractperforms well. Since WFATE is easy for implementation, it will be useful in praticalnetworks. The research work in the thesis covers every steps in the deployment of traf?c en-gineering, and the traf?c mapping technology is especially deeply studied. The resultsand conclusions will be useful in the future theory development and practical deploy-ment of traf?c engineering.