Power Management Policy for Heterogeneous Data Center Based on Histogram and Discrete-Time MDP

Abstract

This work presents a stochastic model for Dynamic Power Management (DPM) that is based on switching-on/off machines in a data center of heterogeneous servers. The aim of a DPM is to ensure both a reasonable energy consumption and an acceptable Quality of Services (QoS). In this paper, arrival jobs and service rates are specified with histograms which are discrete distributions obtained from real traces, empirical data, or incoming traffic measurements. A data center is modeled by a queue, then we formulate the optimization problem by a discrete time Markov Decision Process (MDP) to find the optimal policy. We prove that the optimal policy is not hysteretic. Our approach was applied and tested for several system parameters over real Google traffic traces, when we performed a comparison between homogeneous and heterogeneous data centers.