Study of the correlation between server resources utilization and energy consumption

The amount of data to be treated in data centers is increasing more and more, which makes an obligation to optimize the use of servers in a way that reduces energy consumption and then minimizes the cost. The study of correlation between server resources and energy consumption is a crucial step and involves well-founded strategies to optimize both the use of resources servers and their energy consumption. There are various tools to measure energy consumption of servers. In our work we apply different workloads on servers and then we measure energy consumption using two methods. Internal sensors (performance counters) which allow extracting resources usage information by using listeners on the sensors or external tools, and energy meters and software to acquire data. Error rate of both methods varies according to the integrated sensors in the machine and the type and accuracy of the power meter used. We study the correlation between the power of CPU server and its utilization as well as the correlation between energy consumption of server and its utilization in term of CPU, Memory, and hard drive.