Effect of energy-saving server scheduling on power consumption for large-scale data centers

Masataka  Kato; Hiroyuki Masuyama; Shoji Kasahara; Yutaka Takahashi

doi:10.3934/jimo.2016.12.667

Article Contents

2016, Volume 12, Issue 2: 667-685. Doi: 10.3934/jimo.2016.12.667

This issue Previous Article Analysis of an M/M/1 queue with vacations and impatience timers which depend on the server's states Next Article Analysis and optimization of a gated polling based spectrum allocation mechanism in cognitive radio networks

Effect of energy-saving server scheduling on power consumption for large-scale data centers

1.
Graduate School of Informatics, Kyoto University, Yoshida-Honmachi, Sakyo-ku, Kyoto 606-8501
2.
Graduate School of Information Science, Nara Institute of Science and Technology, 8916-5 Takayama, Ikoma, Nara 630-0192

Received: September 30, 2014

Revised: February 28, 2015

Published: March 2016

Abstract Related Papers Cited by

Abstract

Large-scale data centers for cloud computing services consist of a number of commodity servers, resulting in a huge amount of power consumption. In order to save power consumption, BEEMR (Berkeley Energy Efficient MapReduce), a MapReduce workload manager, is proposed. In a BEEMR-based data center, servers are allocated to either of the interactive and batch zones. Arriving jobs of a small size begin to be processed immediately in the interactive zone, while large-sized jobs are queued and served simultaneously at every fixed service period in the batch zone. In this paper, we analyze the performance of BEEMR-type job scheduling. We consider two queueing models for the interactive and batch zones. The interactive zone is modeled as a single-server queueing system with processor-sharing (PS) service. In terms of the batch zone, we consider a queueing system with gated service in which arriving jobs are queued and begin to be served when a fixed service period starts. For these models, the time-average power consumption and the mean response time are derived. Numerical examples show that the power consumption is significantly affected by the allocation of servers to both zones, while the power consumption is insensitive to the length of the batch-service period.

Keywords:

Mathematics Subject Classification: Primary: 60K25, 60J27; Secondary: 68M20.

Citation:

Related Papers

Cited by

References

[1]	S. Asmussen, Applied Probability and Queues, $2^{nd}$ edition, Springer, New York, 2003.
[2]	K. E. Avrachenkov, U. Ayesta, P. Brown and R. Núñez-Queija, Discriminatory processor sharing revisited, in Proc. IEEE INFOCOM'05, (2005), 784-795.doi: 10.1109/INFCOM.2005.1498310.
[3]	L. A. Barroso and U. Hölzle, The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale Machines, Morgan & Claypool, California, 2009.doi: 10.2200/S00193ED1V01Y200905CAC006.
[4]	P. Bremaud, Markov Chains, Gibbs Fields, Monte Carlo Simulation, and Queues, Springer, New York, 1999.doi: 10.1007/978-1-4757-3124-8.
[5]	L. Breuer and D. Baum, An Introduction to Queueing Theory and Matrix-Analytic Methods, Springer, New York, 2005.
[6]	R. Brown, E. Masanet, B. Nordman, W. Tschudi, A. Shehabi, J. Stanley, J. Koomey, D. Sartor, P. Chan, J. Loper, S. Capana, B. Hedman, R. Duff, E. Haines, D. Sass and A. Fanara, Report to congress on server and data center energy efficiency: Public law 109-431, Lawrence Berkeley National Laboratory, LBNL-363E, 2007.
[7]	Y. Chen, S. Alspaugh, D. Borthakur and R. Katz, Energy efficiency for large-scale MapReduce workloads with significant interactive analysis, in Proc. The European Professional Society on Computer Systems 2012, (2012), 43-56.doi: 10.1145/2168836.2168842.
[8]	D. Gibson and E. Seneta, Monotone infinite stochastic matrices and their augmented truncations, Stochastic Processes and their Applications, 24 (1987), 287-292.doi: 10.1016/0304-4149(87)90019-6.
[9]	H. Masuyama, Error bounds for augmented truncations of discrete-time block-monotone Markov chains under geometric drift conditions, Accepted for publication in Advances in Applied Probability, available online at: http://infosys.sys.i.kyoto-u.ac.jp/~masuyama/papers/ap14504.pdf. doi: 10.1239/aap/1427814582.
[10]	S. Pelley, D. Meisner, T. F. Wenisch and J. W. VanGilder, Understanding and abstracting total data center power, in Proc. Workshop on Energy-Efficient Design 2009, (2009).
[11]	M. Sakata, S. Noguchi and J. Oizumi, An analysis of the M/G/1 queue under round-robin scheduling, Operations Research, 19 (1971), 371-385.
[12]	C. Schwarts, R. Pries and P. Tran-Gia, A queuing analysis of an energy-saving mechanism in data centers, in Proc. International Conference on Information Networking 2012, (2012), 70-75.doi: 10.1109/ICOIN.2012.6164352.
[13]	R. L. Tweedie, Truncation approximations of invariant measures for Markov chains, Journal of Applied Probability, 35 (1998), 517-536.doi: 10.1239/jap/1032265201.
[14]	R. W. Wolff, Stochastic Modeling and the Theory of Queues, Prentice-hall, Englewood Cliffs, NJ, 1989.