Using Thermal-Aware VM Migration Mechanism for High-Availability Cloud Computing

Abstract

Cloud computing, such as Infrastructure as a Service (IaaS), enables vendors to use virtualization technology to rent computing resources on a physical machine to execute the desired applications of users. IaaS is the most common business model of cloud computing; however, its availability remains a concern among users. Several factors affect the availability of a cloud computing center, such as interruption of service caused by hardware component damage. In this study, we focused on the thermal emergency event of CPU overheating caused by chassis fan damage, and determined a method to resolve the crisis before a crash occurs. We designed a thermal-aware VM migration manager (TAVMM) that can determine the health of a physical machine from its temperature and resource use information. By leveraging VM migration, the risk to the physical machine can be removed by transferring its load to a normal one and reducing the CPU temperature. We propose heat transfer and migration time as criteria for a VM selection policy and the load balance algorithm regarding thermal tolerance as the VM allocation policy. The simulation results show that a TAVMM with the proposed VM selection and allocation policy can enhance system ability and reduce the number of VM failures.