Smart temperature monitoring for data center energy efficiency

Energy consumption is a critical concern in data centers as extremely huge amount of energy is consumed every year, a major part of which is used by cooling system. Though cooling system consumes large amount of energy, data centers lack the knowledge of whether energy is efficiently used or not. Recently temperature sensors are increasingly deployed to monitor data center temperature, but where to deploy them is still a tough problem. This paper focuses on optimizing temperature sensor deployment in data centers so as to finely control cooling supply while avoiding temperature alarms at the same time. We solve the problem in two steps. Firstly, assume a temperature snapshot which describes temperature distribution in a rack at a certain point of time is given, we formulate an optimization problem and solve it to find out optimal temperatures by which to adapt CRAC cooling supply. Then, through large amount of optimal temperature data which are selected at different points of time and different work load, we learn the relationship between optimal temperatures and optimal sensor positions. Extensive simulation results show that, compared to other solutions, our mechanism can find better positions for temperature sensors. By deploying temperature sensors on these positions, no matter how device workload changes, cooling supply can be adapted efficiently without causing hot and cold spots.