A Novel DPM Algorithm Based on the Hurst Probability
The key problem of DPM algorithm lies in how to predict the idle period accurately,we proposed a novel Hurst parameter directed probabilistic dynamic power management algorithm that can be applied for embedded systems with multiple power states. We computed the Hurst parameter by analyzing the system workloads self similar degree. Then the Hurst parameter was used to decide the next idle period lengths probability density,which was be used to compute the timeout values that used for controlling power states. Experimental results showed that the Hurst probability based dynamic power management algorithms could save about 80％ energy on the HP hard disk data sets in comparison with the classical method based on probability histogram.
Tao Pin Kong Fei Yang ShiQiang
Computer Science and Technology Department,Tsinghua University Key Laboratory of Pervasive Computing,Ministry of Education,Tsinghua University Beijing 100084,China
国际会议
成都
英文
175-180
2008-01-01（万方平台首次上网日期，不代表论文的发表时间）