Research On LSP Preemption Algorithm For DS-TE Networks

Traditional IP network provides Best-Effort service, so it couldn’t guarantee theQoS. With the rapid development of the Internet, and the increasing varieties ofreal-time service and multi-media applications, the requirement of QoS becamestricter. The present QoS mechanism can’t meet the need, so it is necessary to makefurther research on it. Based on DiffServ-aware MPLS networks, this thesis makes adeep research on how to guarantee the QoS, which mainly focuses on theimprovement and analysis for preemption algorithms, the main work is as follows:Firstly, this thesis describes the network structure, working principle and coretechniques of MPLS technology. And then, it analyzes and compares several majorQoS mechanisms, such as Best-Effort, IntServ and DiffServ model. TheDiffServ-aware traffic engineering technique combines the MPLS TE with DiffServmodel, so it makes use of the explicit route ability of MPLS and the scalability ofDiffServ model, which is effective to solve QoS problems in backbone networks.Secondly, this paper mainly studies the preemption mechanism in DS-TE model,and focuses on current preemption algorithms. According to the defects of V-PREPTalgorithm, the paper proposed two V-PREPT improved algorithms: N-PREPTalgorithm and BH-PREPT algorithm.The improvement of the N-PREPT algorithm is as follows:(1) The executingorder of optimization criterion flexibly changes according to the specificconfiguration of the network weights, and executes the preemption of LSP in varioussituations.(2) Under the premise of minimizing the total cost of preemption,minimizing the number of preempted LSP is considered at first.(3) Multi-layerconstraints are set to make the output of the algorithm more ideal, namely to raiserestriction to the number of preemption.(4)The optimization model and mathematicalexpression are improved so that the cost of minimizing the number of preempted LSPand the waste of bandwidth by the execution of the preemption can be calculatedmore accurately. By MATLAB simulation, the N-PREPT algorithm has betterperformance than the original algorithm in reducing the overhead to the network by rerouting.The improvement of the BH-PREPT algorithm is as follows:(1) Theoptimization criterion of avoiding waste of bandwidth is put in the first place.(2)TheLSP collection which can be preempted is divided into two groups according to thesize of the reserved bandwidth. The combination with the minimum total cost ofpreempted LSP is picked up by taking the enumeration method.(3) The combinationwith the minimum number of preempted LSP is picked up in priority.(4) The adoptedoptimization model and mathematical expression are the same with the N-PREPTalgorithm. Compared with the original algorithm, the BH-PREPT algorithm has betterperformance under various configurations of the network weights.