A performance and power consumption analysis based on processor power models

Recent years witnessed an explosive increase of power consumption in data centers. This increase raised serious concerns due to its high economic and environmental impact. The most important power consumers in data centers are IT hardware and cooling systems. For modern data centers, it is important to select the appropriate power models for improving energy efficiency. This paper provides an overview of recent processor power models, more commonly used for data centers and an analysis of the performance and power consumption for different systems with various types of processors and workloads. In order to perform this analysis, a database was built starting from a data collection from the Standard Performance Evaluation Corporation (SPEC). The data collection was developed with the Server Efficiency Rating Tool (SERT) methodology. SERT has been recognized for several years as one of the best methods for characterizing server efficiency. For the analysis of systems performance and power consumption with various types of processors, the Pearson correlation is then calculated. Processor power models help data centers managers to identify opportunities for power consumption optimization. Their use increases the efficiency of power consumption management and control at a data center level.