An optimized control strategy for load balancing based on live migration of virtual machine

The migration technology of virtual machine has received extensive attention in load balancing. In this paper, we purpose an optimized control strategy which combines multi-strategy mechanism with the prediction mechanism. According to the weighted average of the CPU, memory, I/O and network bandwidth utilization, we divide the hosts into four status domains. The hosts within different load status domain adopt different migration strategy, which considers migration timing, migration candidate VM and migration destination. Through this strategy, it reduces the number of the overloaded hosts, avoids instantaneous peak problem caused by the migration of virtual machines, solves the imbalance problem and the high-cost problem in tradition scheduling algorithm of migration. Experimental results demonstrate that this strategy is able to reduce the migration times and improve system performance. © 2011 IEEE.