HP Labs Technical Reports
Click here for full text:
Idleness is Not Sloth
Golding, Richard; Bosch, Peter; Wilkes, John
Keyword(s):idleness, storage systems, detecting idle period, predicting idle periods
Abstract: Many people have observed that computer systems spend much of their time idle, and various schemes have been proposed to use this idle time productively. We have used this approach to improve overall performance in storage systems. The most common approach is to off- load activity form busy periods to less-busy ones in order to improve system responsiveness. In addition, speculative work can be performed in idle periods in the hope that it will be needed later at times of higher utilization, or a non-renewable resource power can be conserved by disabling unused resources during idle periods. Much existing work in scheduling for idle periods uses ad hoc mechanisms. The work presented here includes a taxonomy of idle-time detection and prediction algorithms that encompasses the prior approaches and also suggests a number of others. We identify metrics that can be used to evaluate all these idle-time detection algorithms and demonstrate the utility of both the metrics and the taxonomy by providing a quantitative evaluation.
Back to Index