Joint resource scheduling and peak power shaving for cloud data centers with distributed Uninterruptible power Supply

Abstract

The grid company enforces high penalties for the peak power demands of cloud data centers. These high penalties result in high electricity bill that can be avoided by relying on the servers' Uninterruptible Power Supply (UPS) as a source of energy during peak load periods. This paper proposes a management framework that exploits the distributed UPS batteries in order to minimize the total cluster's electricity bill. Our framework consists of: i) a scheduler that accounts for both the amount of stored energy and the available resource slacks when making workload placement decision, and ii) a power distributor that decides which UPS battery should store energy and by how much in order to increase the amount of energy that can be accessible for shaving the peak during high-demand periods. Several evaluations based on real Google traces show that our proposed framework achieves significant monetary savings.