GreenSleep: A Multi-sleep Modes Based Scheduling of Servers for Cloud Data Center

Abstract

Due to the high power consumption of data centers, cloud service providers have to suffer the enormous electricity bill as well as environmental problem incurred. In order to solve this problem, huge amount of efforts have been devoted to improving the resource utilization of data center by consolidating requests and switching idle servers into low-power sleeping states. In this paper, we propose GreenSleep, a novel scheduling scheme of servers with multi-sleep modes for cloud data center. Note that, the server under deeper sleep mode usually consumes less power, but needs more time to wake up. To be practical, we assume the incoming requests have multi-dimensional resource demand with different start time and end time. The biggest challenges are: 1) how to allocate the requests to as few servers as possible within execution time; 2) when and which servers should sleep under which mode; 3) when and which sleeping servers should be waken up online. We formulate our problem as an optimization problem with the goal of minimizing the total power consumption of cloud data center. To solve this problem, our algorithm is designed considering both requests allocation and sleep-modes switching. Experiments show that our method can significantly reduce the total power consumption for cloud data center.