Workload management in distributed data centers: thermal and spatial awareness

Abstract

Energy consumption of distributed data centers costs huge amounts to the cloud service providers (CSP) each year. in order to make the services affordable for the cloud service consumers (CSC) and increase the profit for the CSP, it is essential to minimize the cooling and operational costs. in this work, we address cost optimization by taking advantage of time and space varying properties of distributed data centers. We consider the state-of-the-art CRaC units which utilize outside air for cooling purposes and propose a cost model based on the outside cooling which takes into account three different temperature ranges for cooling purpose and operations of CRaC units. Based on the mathematical model, we propose three different heuristics, namely TempCP, TempCD and TempPCD to manage workload using the variation of electricity price, locational outside and within the data center temperature, where the aim is to schedule the incoming request with minimum SLa violations, cooling cost and energy consumption. Experiments conducted on CloudSim platform show that our proposed scheduling policies save between 68% to 75% in terms of cooling cost and between 22% and 39% of overall operational cost compared to the existing scheduling policies.