Dynamic Hybrid-copy Live Virtual Machine Migration: Analysis and Comparison

Abstract

Abstract Live VM migration has significantly gained attention as one of the most key strategies of improving data center efficiency through better resources allocation that leads to efficient services. The data center efficiency includes less power consumption, less unutilized resources, and less service interruption time in cases of maintenance or failures. The key point of the live VM migration is to copy the VM’s state, memory, and disk with the minimal downtime to guarantee less interruption in the services the VM provides. Pre-copy, Post-copy, and Hybrid-copy are examples of Live VM migration methods that copy the VM’s memory states to the destination host, while the VM still runs at the source host, as in the Pre-copy, or at the destination host, as in Post-copy. Each of these classical live migration methods has different characteristics and performance. Dynamic Hybrid-copy combines Pre-copy iterations with the Post-copy to have the best of both methods. In this paper, we present the mathematical model for the Dynamic Hybrid-copy live migration method for the total migration time, downtime, and total transferred data during the migration process. We compare the Dynamic Hybrid-copy with other classical live migration methods. We implement the models on JAVA and evaluate the performance of the Dynamic Hybrid-copy and compare it with other methods for different experimental parameters. Based on the performance results, we show how choosing the effective number of Pre-copy iterations leads to better migration performance. Mathematical models were also implemented in CloudSim and a simulation scenario was tested for the four migration methods.