Optimization of virtual machine placement based on constrained immune memory and immunodominance clone in IaaS cloud mode equipment training

Abstract

In infrastructure as a service (IaaS) cloud mode equipment simulated training, to keep the resource utilization ratio in a rational high level, improve the training effect and reduce the system running cost, the problem of training virtual machine (TVM) placement needs to be resolved first. We make analysis to the problem and give the mathematical formulation to the problem. Then, we figure out the principle and target of the TVM placement. Based on above analysis, we propose a constrained immune memory and immunodominance clone (CIMIC) TVM placement optimization algorithm. By reverse optimization of the initial antibody population, the searching range is reduced. The common antibody population and the immunodominance antibody population evolve simultaneously, which realizes the simultaneous progressing of global searching and local searching of solutions. Further, local optimal is avoided by this means. Memory antibody makes ful use of the unfeasible solutions and the diversity of antibody population is maintained. The constraint information of the problem is utilized to improve the optimization effect. Experiment results show that the CIMIC algorithm improves the overall optimization effect of TVM placement, reduces the server number and improves the resource utilization and system stability.