Efficient multistage bandwidth allocation technique for virtual machine migration in cloud computing

In cloud computing datacenter infrastructure, virtualization has enabled service providers to create abstraction of their physical resources and increased the infrastructure utilization. At the datacenter, cloud users services are implemented using Virtual Machines (VMs). With the popularity of cloud paradigm, numbers of enterprise applications deployed over cloud systems are increasing. This shift in cloud paradigm has brought newchallenges of server consolidation, load balancing, resource management, and server fault situation. Cloud managers trigger VM migration to achieve all these challenges. VM migration allows migration of running VMs from one datacenter to another. In migration process to keep the VM alive, VM memory pages are transferred in multiple iterations. Thus, there is an increase in the requirement of network resources consumption during migration process. However, inaccurate bandwidth allocation for a VM migration request can lead to performance degradation. Therefore, there is a need to devise an efficient bandwidth allocation strategy. In this paper, we propose a multistage bandwidth allocation technique which aims to minimize the bandwidth allocation at each round of migration mechanism. The proposed technique is implemented by the simulation carried out in Matlab software. To ensure the applicability for real-time hosted applications, multistage scheme is evaluated using both low and high dirty rate conditions. We also considered distributed cloud deployment situation by using sequential and multiple VM migration test cases. The results demonstrated shows, as compared to existing VM migration bandwidth allocation schemes, our proposed technique outperform in terms of downtime, migration time, bandwidth provisioned and migration iterations count. Additionally, we also explored multistage technique with fuzzy logic to improve the VM selection process. The significance of this study is that, multistage technique allows bandwidth allocation in each iteration as variable to current dirty rate scenario, thereby increasing the revenue of cloud providers.