A power and thermal-aware virtual machine allocation mechanism for Cloud data centers

Abstract

With the rapid growing number of Cloud applications, demands for large-scale data centers have raised to historical high. Cloud data centers allow dynamic and flexible resource provisioning to accommodate time varying computational demands. Recent studies have proposed several allocation policies based mainly on power consumption of servers. Host temperature, however, is rarely considered as a monitoring parameter. This work proposes a power and thermal-aware virtual machine (VM) allocation mechanism for Cloud data centers. The objective of the proposed mechanism is to reduce the overall energy consumption and VM migration numbers, while avoiding violations of Service Level Agreements (SLA) in Cloud data centers. The proposed mechanism was implemented and evaluated on CloudSim. Simulation results show that the proposed allocation mechanism brings significant benefits in terms of energy saving and other performance indices.