Understanding Desktop Energy Footprint in an Academic Computer Lab

In this paper, we present results from our measurement and analysis of energy use and user behavior in an academic computer lab. We use wireless power sensors to collect power readings from 22 computers in the lab. We use software monitor to log various user activities in the computer. We collected a total of 59.6 million power readings and 220.3 million user activity logs over one month. We analyze the data collected from this instrumentation to not only understand how much energy is used but to also drill down and reveal a detailed understanding of which machines, processes, and users contribute the most to computing energy footprint of the lab. Our results show the power draw on different machines in the lab are different despite the identical hardware and software settings. Our study attributes this difference to different users presenting different types and lengths of load to the computers, preferring specific physical computers, which leads to some computer being used more than others. The results show that the majority of energy was wasted while the computer was left in idle mode, and individual user behavior affects the energy consumption.