Achieving Energy Efficiency in Datacenters by Virtual Machine Sizing, Replication, and Placement

Monitoring and analyzing virtual machines (VMs) in a datacenter enables a cloud provider to place them on fewer physical machines with negligible performance penalty, a process which is known as the VM consolidation. Such a consolidation allows the cloud provider to take advantage of dissimilar workloads to reduce the number of ON servers, and thereby, reduce the datacenter energy consumption. Moreover, placing multiple copies of VMs on different servers and distributing the incoming requests among them can reduce the resource requirement for each copy and help the cloud provider do more aggressive consolidation. This chapter begins by a substantial review of various approaches for consolidation, resource management, and power control in datacenters. It continues by presenting a dynamic programming-based algorithm for creating multiple copies of a VM without degrading performance and doing VM consolidation for the purpose of datacenter energy minimization. A side benefit of the consolidation is to improve reliability of the services provided by the datacenter to its clients. Using the proposed algorithm, it is shown that more than 20% energy saving can be achieved compared to the previous work.