Event-Based Power Estimation

The increasing awareness of power consumption in computer systems focuses attention on the subject of measuring the actual power consumed at any given time. While performance measurement and monitoring is directly supported at the architecture level, and implemented in various software tools, power measurement and monitoring is lagging behind. A software-only power estimation capability could be useful for power management policies, power-aware code optimizations, and other power-related administrative tasks.

The system reported here uses a calibrated power model to get accurate power estimations, without any external devices, including the breakdown of power usage into major power-consuming blocks. Calibration is accomplished using multiple linear regression over a predefined dataset.

The power model allows the breakdown of the estimated power to the consumers associated with each event, thereby providing important insight regarding where power is spent in the application.

The estimation error of this model for a subset of the SPEC CPU2000 benchmark (the full integer suite, plus three of the FP suite) is less then 0.25% for overall energy consumption, and less then 2% for online power consumption.

By: Yaakov Yaari

Published in: H-0191 in 2003

LIMITED DISTRIBUTION NOTICE:

This Research Report is available. This report has been submitted for publication outside of IBM and will probably be copyrighted if accepted for publication. It has been issued as a Research Report for early dissemination of its contents. In view of the transfer of copyright to the outside publisher, its distribution outside of IBM prior to publication should be limited to peer communications and specific requests. After outside publication, requests should be filled only by reprints or legally obtained copies of the article (e.g., payment of royalties). I have read and understand this notice and am a member of the scientific community outside or inside of IBM seeking a single copy only.

H-0191.pdf

Questions about this service can be mailed to reports@us.ibm.com .