Thermal-Aware Performance Optimization in Power Constrained Heterogenous Data Centers

The power consumption of data centers has been increasing at a rapid rate over the past few years. Many of these data centers experience physical limitations on the power needed to run the data center. This paper attempts to maximize the performance of a data center that is subject to total power consumption and thermal constraints. We consider a power model for a data center that includes power consumed in both Computer Room Air Conditioning (CRAC) units and compute nodes. Our approach quantifies the performance of the data center as the total reward collected from completing tasks in a workload by their individual deadlines. We develop novel optimization techniques for assigning the performance states of compute cores at the data center level to increase the performance of the data center. The assignment problem in this paper is thermal aware as it considers the temperature evolution effects of performance state assignments, which in turn affects the power consumed by the CRAC units. Our simulation studies show that in some cases the assignment technique used in this paper achieves about 10% average improvement in the performance of a data center over an assignment problem that only considers putting a compute core in the performance state with the highest performance or turning the core off.