2010 Poster Sessions : Identifying Energy Waste through Power and Utilization Monitoring

Student Name : Maria Kazandjieva
Advisor : Philip Levis
Research Areas: Computer Systems
Abstract:
Computing systems are a common part of most enterprise building and their power consumption is non-negligible. According to the DoE, computing systems in office and education buildings account for about 1.5% of the annual US electricity consumption. Our Computer Science building, for example, spends close to half a million dollars a year for electricity.
The PowerNet project monitors power consumption of 140 different devices in the Gates Computer Science building. In addition, we collect utilization statistics, CPU and network traffic, that help understand the workloads that the computing infrastructure is supporting. Using power and utilization data collected over the last 3 to 14 months, we can create a breakdown of power consumption. We find that computing accounts for ~60% of total building energy budget.
Furthermore, a large amount this energy is wasted because computers and networks are heavily under-utilized. The data analysis suggests that there are many opportunities for reducing consumption and cost by rethinking the way we design building-scale computing systems.

Bio:
Maria Kazandjieva is a 3rd year PhD at Stanford University's Computer Science Department. Her research interests include energy, sensing, data analysis. Before Stanford, Maria worked at Princeton University as a research assistant. She holds a BA from Mount Holyoke College with major in Computer Science and minor in Mathematics.