A Performance-Guaranteed Virtual Machine Migration Strategy in Cloud

Abstract

In cloud computing, frequent addition and removal of virtual machines (VMs) can cause the resource waste or load imbalance among physical machines (PMs). Replacement of VMs to appropriate PMs needs to be performed periodically, which is called VM migration (VMM). The exist VMM methods mainly aim at optimizing resource utilization for cloud providers, but they neglect the VM performance, which may affect user experience. In contract to existing VMM methods, this paper proposes a Performance-Guaranteed Virtual Machine Migration Strategy (PGVMM) from users' perspectives, which tries to guarantee VM performance for users. First, we study the relationship between the PM workload and the VM performance to improve our previous VM performance model, which can offer help to the following VMM. Second, we take the VM performance and migration cost into account to formulate the VMM as an optimization problem, which tries to guarantee VMs performance with the least migration cost. Third, we propose a dual-threshold greedy-based algorithm to solve the VMM problem efficiently. Then some experiments are conducted to evaluate our algorithm. The results also prove the efficiency of the proposed method.