poweraware-15 - Dynamic Power Management Using Adaptive...

Info iconThis preview shows pages 1–2. Sign up to view the full content.

View Full Document Right Arrow Icon
Dynamic Power Management Using Adaptive Learning Tree Eui-Young Chung Luca Benini Giovanni De Micheli eychung,nanni @stanford.edu Stanford University Computer Systems Laboratory Stanford, CA 94305-4070, USA lbenini@deis.unibo.it Universit`a di Bologna Dip. Informatica, Elettronica, Sistemistica 40136, Bologna, ITALY Abstract Dynamic Power Management (DPM) is a technique to reduce power consumption of electronic systems by selectively shutting down idle components. The quality of the shutdown control algo- rithm (power management policy) mostly depends on the knowl- edge of user behavior, which in many cases is initially unknown or non-stationary. For this reason, DPM policies should be ca- pable of adapting to changes in user behavior. In this paper, we present a novel DPM scheme based on idle period clustering and adaptive learning trees. We also provide a design guide for ap- plying our technique to components with multiple sleep states. Experimental results show that our technique outperforms other advanced DPM schemes as well as simple time-out policies. The proposed approach shows little deviation of efficiency for various workloads having different characteristics, while other policies show that their efficiency changes drastically depending on the trace data characteristics. Furthermore, experimental evidence indicates that our workload learning algorithm is stable and has fast convergence. 1 Introduction The importance of system-level low-power design techniques has been increased by the widespread use of portable devices, which have limited battery life time [2, 4, 7]. Dynamic power management ( DPM ) [1] is a system-level low power design tech- nique aiming at controlling performance and power levels of dig- ital circuits and systems, by exploiting the idleness of their com- ponents. The heart of DPM is a Power Manager ( PM )wh ich monitors the overall system state and issues commands to con- trol the power state of the system when it detects idleness. The control algorithm implemented by the PM is called a power man- agement policy . Adaptivity is one of the most important issues in DPM because most external environments (user requests) are non-stationary. Three classes of power management policies have been pro- posed in the past: time-out, predictive, and stochastic policies. The fixed time-out policy shuts down the system after a fixed amount of idle time [15]. Adaptive time-out policies are more efficient because they change the time-out according to the pre- vious history. In contrast with time-out policies, predictive tech- niques do not wait for a time-out to expire, but shut down the system as soon as it becomes idle if they predict that the idle time will be long enough to amortize the cost of shutting down.
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 2
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 09/01/2009 for the course CSE CS-699 taught by Professor Prf.p.bhaduri during the Spring '09 term at Indian Institute of Technology, Guwahati.

Page1 / 6

poweraware-15 - Dynamic Power Management Using Adaptive...

This preview shows document pages 1 - 2. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online