Thermal aware workload consolidation in cloud data centers

Abstract

This paper addresses the problem of Data Centers (DC) energy efficiency by proposing a complex consolidation technique which aims at optimizing the DC operation such that both the IT equipment power consumption and cooling system power consumption is minimized. The paper presents a thermo-electrical DC model that defines the relation between the server utilization due to VM deployment, the server room ambient temperature and the cooling system load needed to dissipate the corresponding heat. Furthermore, a set of approximation algorithms and heuristics are presented to solve the resulting NP-hard thermal-aware consolidation problem. Finally, simulation results on some predefined scenarios show that the proposed techniques improve the DC energy consumption with 5% up to 20% compared to the well-known First-Fit algorithm by deploying VMs on servers with low thermal influence. This allows the cooling system to increase its air supply temperature with up to 10 degrees Celsius, leading to lower power consumption.