Vm Migration Planning In Software-defined Data Center Networks

Modern data centers exploit virtualization,which allows multiple independent virtual machines(VMs)to run on the same physical server,for efficient resource management to reduce cost and energy budget.For achieving various network optimization and management goals(e.g.,energy saving,load balancing)in data centers,some VMs need to be migrated from the initial servers to the target ones.Live VM migration is an effective technique to achieve these goals.When multiple VMs need to be migrated at the same time,the migration order and migration bandwidth has a great impact on migration performance.With the limited network bandwidth resources,bandwidth allocation scheme is also an important influence factor.Therefore,it is necessary to make an efficient VM migration planning for determining migration order and migration bandwidth to improve VM migration performance.The scale of data centers expand rapidly.However,the traditional network technology seriously reduces the degree of virtualization.The emergence of Software Defined Networking(SDN)can effectively solve this problem,and provide an effective solution for VM migration.SDN global control logic can fully meet the multi-path routing and other requirements.SDN can increase the transmission capacity of network which can reduce the total migration time and downtime of VMs.Therefore,in software-defined data center networks,for making an effective VM migration planning to improve the migration performance,we design the Individual Performance Guaranteed VM Migration(IPG-VMM)algorithm and PSO-based VM Migration(PSO-VMM)algorithm.(1)IPG-VMM algorithm can maximize the migration bandwidth with the limited network bandwidth resources by preprocessing and minimizing the correlations between migration schemes of different VMs while guaranteeing the migration performance of individual VMs.Meanwhile,IPG-VMM algorithm can solve the "idle bandwidth" problem in the step-by-step migration model and make full use of network bandwidth to improve network transmission capacity.(2)To get a better VM migration planning,we use swarm intelligence algorithm with some time cost.This problem is a combinatorial optimization problem,and the Particle Swarm Optimization(PSO)is a powerful algorithm to solve combinatorial optimization problems and the performance of PSO algorithm is proved.Therefore,we improve the updating of particles in PSO algorithm by merging particles and propose the PSO-VMM algorithm.(3)Extensive experiments show that IPG-VMM algorithm can reduce the total migration time and total downtime by up to 25%(resp.65%)and 50%(resp.30%)compared with the FPTA(resp.CQNCR)algorithm,respectively.With some time cost PSO-VMM algorithm is better than IPG-VMM algorithm.