Efficient datacenter resource utilization through cloud resource overcommitment

We propose an efficient resource allocation framework for overcommitted clouds that makes great energy savings by 1) minimizing PM overloads via resource usage prediction, and 2) reducing the number of active PMs via efficient VM placement and migration. Using real Google traces collected from a cluster containing more than 12K PMs, we show that our proposed techniques outperform existing ones by minimizing migration overhead, increasing resource utilization, and reducing energy consumption.