Energy consumption is a growing issue in data centers, impacting theireconomic viability and their public image. In this work we empiricallycharacterize the power and energy consumed by different types of servers. Inparticular, in order to understand the behavior of their energy and powerconsumption, we perform measurements in different servers. In each of them, weexhaustively measure the power consumed by the CPU, the disk, and the networkinterface under different configurations, identifying the optimal operationallevels. One interesting conclusion of our study is that the curve that definesthe minimal CPU power as a function of the load is neither linear nor purelyconvex as has been previously assumed. Moreover, we find that the efficiency ofthe various server components can be maximized by tuning the CPU frequency andthe number of active cores as a function of the system and network load, whilethe block size of I/O operations should be always maximized by applications. Wealso show how to estimate the energy consumed by an application as a functionof some simple parameters, like the CPU load, and the disk and networkactivity. We validate the proposed approach by accurately estimating the energyof a map-reduce computation in a Hadoop platform.
展开▼
