Virtual machine migration techniques for managing time-varied workloads

This paper studies the virtual machine migration problem under dynamic traffic environment in data centers, while considering the tradeoff among resource utilization, energy consumption, and migration frequency. We use the service level agreement (SLA) soft threshold model to avoid frequent peak migrations at the migration trigger point. In the source virtual machine selection process, the maximum relevance is considered during the virtual machines selection method to improve resource coupling. In the destination server selection, the multi-resource relevance and matching method is used to solve the coupling problem of a single resource and the matching problem of a variety of resources. The simulation results show the proposed algorithm can improve the utilization rate of server resources to reduce power consumption and ensure system stability by reducing migration times. We also show that compared to existing techniques the algorithm used in this study can better adapt to dynamic changes of workloads.