New power consumption prediction procedure using performance counters to increase computer power supply efficiency

Abstract:
As the number of data centers has grown in pace with the burgeoning expansion in information technology (IT), reducing the power consumption of data centers has become an important social challenge. IT server equipment is a primary contributor to data center power consumption and, as such, is an area where power conservation is indispensable. Nevertheless, while the efficiency of the power supply modules integrated into servers has recently seen significant improvements, their overall efficiency generally depends on load rates. This is especially true under low power load conditions, where it is known that efficiency decreases drastically. Recently, power-saving techniques that work by controlling the power module configuration under low power load conditions have been the subject of increased research, and based on such techniques, further efficiency improvements can be expected by interlocking the real-time server load status with the power supply configuration control. In this study, the performance counters built into the processor of a server are used to predict server power load variations and a function that predicts server power consumption levels is defined. In a server application experiment utilizing prototype computer hardware and multiple regression analysis, we validated that the function could precisely predict server power requirements. Furthermore, our experimental results show that power supply efficiency improvements up to 14% could be achieved using our model equations to control the efficiency of power supply modules.