Dynamic Power Management Using Machine Learning

Dynamic power management (DPM) work proposed to date places inactive components into low power states using a single DPM policy. In contrast, we instead dynamically select among a set of DPM policies with a machine learning algorithm. We leverage the fact that different policies outperform each other under different workloads and devices. Our algorithm adapts to changes in workloads and guarantees quick convergence to the best performing policy for each workload. We performed experiments with a policy set representing state of the art DPM policies on a hard disk drive and a WLAN card. Our results show that our algorithm adapts really well with changing device and workload characteristics and achieves an overall performance comparable to the best performing policy at any point of time

[1]  S. McFarling.  Combining Branch Predictors , 1993 .

[2]  Yale N. Patt,et al.  Alternative implementations of hybrid branch predictors , 1995, MICRO 1995.

[3]  L. Benini,et al.  Dynamic power management of laptop hard disk , 2000, Proceedings Design, Automation and Test in Europe Conference and Exhibition 2000 (Cat. No. PR00537).

[4]  Allen C.-H. Wu,et al.  A predictive system shutdown method for energy saving of event-driven computation , 1997, 1997 Proceedings of IEEE International Conference on Computer Aided Design (ICCAD).

[5]  Massoud Pedram,et al.  Dynamic power management based on continuous-time Markov decision processes , 1999, DAC '99.

[6]  Yale N. Patt,et al.  An effective programmable prefetch engine for on-chip caches , 1995, MICRO 1995.

[7]  Mani B. Srivastava,et al.  Predictive system shutdown and other architectural techniques for energy efficient programmable computation , 1996, IEEE Trans. Very Large Scale Integr. Syst..

[8]  Carl Staelin,et al.  Idleness is Not Sloth , 1995, USENIX.

[9]  Luca Benini,et al.  Dynamic power management using adaptive learning tree , 1999, 1999 IEEE/ACM International Conference on Computer-Aided Design. Digest of Technical Papers (Cat. No.99CH37051).

[10]  Luca Benini,et al.  Event-driven power management of portable systems , 1999, Proceedings 12th International Symposium on System Synthesis.

[11]  Radu Marculescu,et al.  Hierarchical adaptive dynamic power management , 2004, Proceedings Design, Automation and Test in Europe Conference and Exhibition.

[12]  Luca Benini,et al.  Quantitative comparison of power management algorithms , 2000, Proceedings Design, Automation and Test in Europe Conference and Exhibition 2000 (Cat. No. PR00537).

[13]  Fred Douglis,et al.  Adaptive Disk Spin-Down Policies for Mobile Computers , 1995, Comput. Syst..

[14]  John Wilkes,et al.  UNIX Disk Access Patterns , 1993, USENIX Winter.

[15]  Yoav Freund,et al.  A decision-theoretic generalization of on-line learning and an application to boosting , 1995, EuroCOLT.

[16]  Luca Benini,et al.  Dynamic power management - design techniques and CAD tools , 1997 .

[17]  Yale N. Patt,et al.  Alternative implementations of hybrid branch predictors , 1995, Proceedings of the 28th Annual International Symposium on Microarchitecture.

[18]  Luca Benini,et al.  Policy optimization for dynamic power management , 1998, Proceedings 1998 Design and Automation Conference. 35th DAC. (Cat. No.98CH36175).

[19]  WuAllen C.-H.,et al.  A predictive system shutdown method for energy saving of event-driven computation , 2000 .

[20]  Anna R. Karlin,et al.  Competitive randomized algorithms for non-uniform problems , 1990, SODA '90.

[21]  Luca Benini,et al.  Dynamic Power Management , 1998 .