Analysis of demand response for datacenter energy management using GA and time-of-use prices

Abstract

This paper presents an energy management algorithm for a grid-connected datacenter to provide cost-efficient consumption opportunity in a Demand-Response (DR) program. The aim of this study is to schedule controllable loads of the datacenter based on on-peak and off-peak periods of different time-varying electricity prices. The electricity consumption of the datacenter is modeled with two types of workload: batch and service jobs. Service jobs are executed as soon as submitted (as non-controllable) while batch jobs can be executed over the time horizon (as controllable). Moreover, the power demand of the batch jobs can be modified by providing less/more processing resources during their execution (degradation mode). Therefore, the batch jobs give flexibility in order to provide participation opportunity to DR program for the datacenter. In this study, we determine the profitability of different time-of-use electricity prices on datacenter by solving cost-minimization problem using genetic algorithm. Simulation results show that the presented decision algorithm is able to reduce the electricity cost by participating in DR without decreasing the quality of services.