Optimization Techniques for Virtual Machine Placement and Migration

Abstract: Virtualization is widely used due to its flexibility, scalability, and cost reduction. In virtualization, virtual machines (VM) should be placed optimally onto physical machines (PM) to reduce power consumption and avoid resource shortages. VM placement is an intractable combinatorial optimization problem. Moreover, optimal VM placement changes if the loads on VMs change over time. This means that load change necessitates VM migration among PMs. Since VM migration incurs network load, migration frequency must be small. Thus, both power consumption and the number of migrations should be minimized when determining VM placement. This research formulates the problem and examines algorithms that solve it. The examined algorithms include two metaheuristics, i.e., simulated annealing and tabu search methods. A method previously presented by the author was also tested for comparison. These methods were evaluated through computer simulation.