A feedback control approach to dynamic speed scaling in computing systems

Speed scaling concerns the dynamic adaptation of the active service capacity of a computing system to the processing demands. This problem has received recent attention, motivated by balancing performance with energy consumption; various proposals have been suggested where the processor speed is a function of the current job population, combined with an appropiate scheduling discipline. In this paper we cast the problem in the setting of feedback control, using a fluid model of the queueing system; in this framework the problem is of designing a controller to track the exogenous demand, and the prior work can be seen as restricting the controller to a static function. By allowing for a dynamic controller, in particular a proportional-integral law, we show how the relevant performance tradeoff can be improved. We further indicate a discrete server implementation of this control law, based on a mix of dedicated servers and pooled helpers; its performance is evaluated analytically and by simulation.

[1]  Lachlan L. H. Andrew,et al.  Power-aware speed scaling in processor sharing systems: Optimality and robustness , 2012, Perform. Evaluation.

[2]  Susanne Albers,et al.  Energy-efficient algorithms for flow time minimization , 2006, STACS.

[3]  Philippe Robert.  Stochastic Networks and Queues , 2003 .

[4]  Lachlan L. H. Andrew,et al.  Optimal speed scaling under arbitrary power functions , 2009, SIGMETRICS Perform. Evaluation Rev..

[5]  E. Yaz,et al.  Linear optimal control, H2 and H∞ methods, by Jeffrey B. Burl, Addison Wesley Longman, Inc. Menlo Park, CA, 1999 , 2000 .

[6]  Liang Zheng,et al.  On the Viability of a Cloud Virtual Service Provider , 2016, SIGMETRICS.

[7]  A. Wierman,et al.  Optimality, fairness, and robustness in speed scaling designs , 2010, SIGMETRICS '10.

[8]  Na Li,et al.  On the Interaction Between Load Balancing and Speed Scaling , 2015, IEEE Journal on Selected Areas in Communications.

[9]  Mor Harchol-Balter.  Performance Modeling and Design of Computer Systems: The M/G/1 Queue and the Inspection Paradox , 2013 .

[10]  Linus Schrage,et al.  Letter to the Editor - A Proof of the Optimality of the Shortest Remaining Processing Time Discipline , 1968, Oper. Res..

[11]  F. Frances Yao,et al.  A scheduling model for reduced CPU energy , 1995, Proceedings of IEEE 36th Annual Foundations of Computer Science.

[12]  J. Michael Harrison,et al.  Dynamic Control of a Queue with Adjustable Service Rate , 2001, Oper. Res..

[13]  Kirk Pruhs,et al.  Getting the Best Response for Your Erg , 2004, SWAT.

[14]  N. Bansal,et al.  Speed scaling with an arbitrary power function , 2009, SODA 2009.