Thermal modeling and management of storage systems in data centers

Abstract

Thermal modeling and management techniques have been widely investigated in recent years. Prior studies show thermal management could increase energy efficiency of data centers. The thermal impacts of CPUs on data storage have been extensively studied; however, disk thermal models are still in their infancy. In our study, we aim at building thermal models that take into account both CPUs and disks. We propose an approach to developing thermal models to estimate a data node’s outlet temperature based on its CPU and disk activities. Integrating our thermal model into an energy consumption model, we can evaluate the total energy cost of a data center. We apply our models to study the impact of various thermal management strategies on energy efficiency.