Set-Based Comprehensive Learning Particle Swarm optimization for Virtual Machine Placement Problem

The virtual machine placement (VMP) is a significant technology in energy-saving field, which is an increasingly important problem of cloud computing centers. Most existing algorithms are difficult to handle the large-scale VMP problems with heterogeneous resources and large demand of virtual machines. In this paper, we propose the set-based comprehensive learning particle swarm optimization (SCLPSO) to solve the VMP problem. SCLPSO combines the set-based particle swarm optimization framework (S-PSO) with the comprehensive learning particle swarm optimizer. With the S - PSO framework, SCLPSO is able to solve the VMP problem which is defined on the discrete search space. With the redefined velocity updating rule in SCLPSO, each dimension of a particle can potentially learn from different exemplars. This strategy improves the exploration of the algorithm. The algorithm also introduces a heuristic factor to guide a virtual machine (VM) to be placed on a more suitable physical machine (PM), which improves the resource utilization of the PM. Based on the devised strategies, large-scale VMP problems with heterogeneous resources can be well resolved by SCLPSO. We conduct experiments on different instances and compare SCLPSO with other classical algorithms. The experimental results demonstrate that the proposed algorithm is promising.