Research On Load Balancing Algorithm Based On Virtual Machine Migration In Cloud Data Center

In recent years, with the rapid development of cloud computing, the related technologies have been widely used in various fields. Virtualization has become an integral part in the design and operation of modern data center, which can realize the isolation between different applications and reduce management complexity. However, virtualization brings not only the above advantages, it can also transfer tasks from heavily loaded servers to lightly loaded servers through virtual machine (VM) migrations to achieve load balancing.Many load balancing algorithms using VM migration have been proposed recently. However, most of the existing algorithms only consider the server-side resource load balancing, but ignore the inter-VM dependencies, which lead to the increase of the traffic bandwidth consumption unconsciously. To solve this problem, this paper proposed a VM migration strategy for minimizing the network traffic, which considered the inter-VM dependencies and the underlying network structure. By optimizing the communication cost and migration cost, it can reduce the network traffic considerably.VMs use multiple resources to serve a variety of applications in cloud data center. And the demand for resources between applications may be different, resulting in different overutilized resources in different servers. So it is necessary to select VMs to migrate out from overloaded host according to the demands of different resources. For example, we should choose the VM with high CPU usage to migrate out from a CPU overutilized host. Otherwise, it will lead to frequent migrations and increase the network traffic eventually. For the above reason, we proposed a traffic and resource intensity aware load balancing algorithm based on the new VM migration strategy. This algorithm not only reduces communication cost and migration cost, but also reduces the time to achieve load balance and avoid future load imbalance which resulting in lower migrations. The simulation results show the superior performance of the new algorithm to other methods.