Profiling Power Consumption on Desktop Computer Systems

Background. Energy awareness in the ICT has become an important issue: ICT is both a key player in energy efficiency, and a power drainer. Focusing on software, recent work suggested the existence of a relationship between power consumption, software configuration and usage patterns in computer systems. Aim. The aim of this work was collecting and analysing power consumption data of a general-purpose computer system, simulating common usage scenarios, in order to extract a power consumption profile for each scenario. Methods. We selected a desktop system running Windows XP as a test machine. Meanwhile, we developed 11 usage scenarios, classified by their functionality, and automated by a GUI testing tool. Then, we conducted several test runs of the scenarios, collecting power consumption data by means of a power meter. Results. Our analysis resulted in an estimation of a power consumption value for each scenario and software application used, obtaining that each single scenario introduced an overhead from 2 to 11 Watts, corresponding to an increase of about 12%. Conclusions. We determined that software and its usage patterns impacts consistently on the power consumption of computer systems. Further work will be devoted to evaluate how power consumption is affected by the usage of specific system resources, like processors, disks, memory etc.