Distributed dynamic speed scaling

Rade Stanojevic, Robert Shorten

Research output: Chapter in Book/Report/Conference proceedingConference contribution

38 Citations (Scopus)

Abstract

In recent years we have witnessed a great interest in large distributed computing platforms, also known as clouds. While these systems offer enormous computing power, they are major energy consumers. In existing data centers CPUs are responsible for approximately half of the energy consumed by the servers. A promising technique for saving CPU energy consumption is dynamic speed scaling, in which the speed at which the processor is run is adjusted based on demand and performance constraints. In this paper we look at the problem of allocating the demand in the network of processors (each being capable to perform dynamic speed scaling) to minimize the global energy consumption/cost subject to a performance constraint. The nonlinear dependence between the energy consumption and the performance as well as the high variability in the energy prices result in a nontrivial resource allocation. The problem can be abstracted as a fully distributed convex optimization with a linear constraint. On the theoretical side, we propose two low-overhead fully decentralized algorithms for solving the problem of interest and provide closed-form conditions that ensure stability of the algorithms. Then we evaluate the efficacy of the optimal solution using simulations driven by the real-world energy prices. Our findings indicate a possible cost reduction of 10- 40% compared to power-oblivious 1/N load balancing, for a wide range of load factors.

Original languageEnglish
Title of host publication2010 Proceedings IEEE INFOCOM
DOIs
Publication statusPublished - 2010
Externally publishedYes
EventIEEE INFOCOM 2010 - San Diego, CA, United States
Duration: 14 Mar 201019 Mar 2010

Other

OtherIEEE INFOCOM 2010
CountryUnited States
CitySan Diego, CA
Period14/3/1019/3/10

Fingerprint

Energy utilization
Resource allocation
Program processors
Convex optimization
Distributed computer systems
Cost reduction
Servers
Costs

Keywords

  • Data center
  • Distributed coordination
  • Dynamic speed scaling
  • Energy management

ASJC Scopus subject areas

  • Computer Science(all)
  • Electrical and Electronic Engineering

Cite this

Stanojevic, R., & Shorten, R. (2010). Distributed dynamic speed scaling. In 2010 Proceedings IEEE INFOCOM [5462197] https://doi.org/10.1109/INFCOM.2010.5462197

Distributed dynamic speed scaling. / Stanojevic, Rade; Shorten, Robert.

2010 Proceedings IEEE INFOCOM. 2010. 5462197.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Stanojevic, R & Shorten, R 2010, Distributed dynamic speed scaling. in 2010 Proceedings IEEE INFOCOM., 5462197, IEEE INFOCOM 2010, San Diego, CA, United States, 14/3/10. https://doi.org/10.1109/INFCOM.2010.5462197
Stanojevic R, Shorten R. Distributed dynamic speed scaling. In 2010 Proceedings IEEE INFOCOM. 2010. 5462197 https://doi.org/10.1109/INFCOM.2010.5462197
Stanojevic, Rade ; Shorten, Robert. / Distributed dynamic speed scaling. 2010 Proceedings IEEE INFOCOM. 2010.
@inproceedings{87dbb82658064b84b2b9efc3c44de1f7,
title = "Distributed dynamic speed scaling",
abstract = "In recent years we have witnessed a great interest in large distributed computing platforms, also known as clouds. While these systems offer enormous computing power, they are major energy consumers. In existing data centers CPUs are responsible for approximately half of the energy consumed by the servers. A promising technique for saving CPU energy consumption is dynamic speed scaling, in which the speed at which the processor is run is adjusted based on demand and performance constraints. In this paper we look at the problem of allocating the demand in the network of processors (each being capable to perform dynamic speed scaling) to minimize the global energy consumption/cost subject to a performance constraint. The nonlinear dependence between the energy consumption and the performance as well as the high variability in the energy prices result in a nontrivial resource allocation. The problem can be abstracted as a fully distributed convex optimization with a linear constraint. On the theoretical side, we propose two low-overhead fully decentralized algorithms for solving the problem of interest and provide closed-form conditions that ensure stability of the algorithms. Then we evaluate the efficacy of the optimal solution using simulations driven by the real-world energy prices. Our findings indicate a possible cost reduction of 10- 40{\%} compared to power-oblivious 1/N load balancing, for a wide range of load factors.",
keywords = "Data center, Distributed coordination, Dynamic speed scaling, Energy management",
author = "Rade Stanojevic and Robert Shorten",
year = "2010",
doi = "10.1109/INFCOM.2010.5462197",
language = "English",
isbn = "9781424458363",
booktitle = "2010 Proceedings IEEE INFOCOM",

}

TY - GEN

T1 - Distributed dynamic speed scaling

AU - Stanojevic, Rade

AU - Shorten, Robert

PY - 2010

Y1 - 2010

N2 - In recent years we have witnessed a great interest in large distributed computing platforms, also known as clouds. While these systems offer enormous computing power, they are major energy consumers. In existing data centers CPUs are responsible for approximately half of the energy consumed by the servers. A promising technique for saving CPU energy consumption is dynamic speed scaling, in which the speed at which the processor is run is adjusted based on demand and performance constraints. In this paper we look at the problem of allocating the demand in the network of processors (each being capable to perform dynamic speed scaling) to minimize the global energy consumption/cost subject to a performance constraint. The nonlinear dependence between the energy consumption and the performance as well as the high variability in the energy prices result in a nontrivial resource allocation. The problem can be abstracted as a fully distributed convex optimization with a linear constraint. On the theoretical side, we propose two low-overhead fully decentralized algorithms for solving the problem of interest and provide closed-form conditions that ensure stability of the algorithms. Then we evaluate the efficacy of the optimal solution using simulations driven by the real-world energy prices. Our findings indicate a possible cost reduction of 10- 40% compared to power-oblivious 1/N load balancing, for a wide range of load factors.

AB - In recent years we have witnessed a great interest in large distributed computing platforms, also known as clouds. While these systems offer enormous computing power, they are major energy consumers. In existing data centers CPUs are responsible for approximately half of the energy consumed by the servers. A promising technique for saving CPU energy consumption is dynamic speed scaling, in which the speed at which the processor is run is adjusted based on demand and performance constraints. In this paper we look at the problem of allocating the demand in the network of processors (each being capable to perform dynamic speed scaling) to minimize the global energy consumption/cost subject to a performance constraint. The nonlinear dependence between the energy consumption and the performance as well as the high variability in the energy prices result in a nontrivial resource allocation. The problem can be abstracted as a fully distributed convex optimization with a linear constraint. On the theoretical side, we propose two low-overhead fully decentralized algorithms for solving the problem of interest and provide closed-form conditions that ensure stability of the algorithms. Then we evaluate the efficacy of the optimal solution using simulations driven by the real-world energy prices. Our findings indicate a possible cost reduction of 10- 40% compared to power-oblivious 1/N load balancing, for a wide range of load factors.

KW - Data center

KW - Distributed coordination

KW - Dynamic speed scaling

KW - Energy management

UR - http://www.scopus.com/inward/record.url?scp=77953294707&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77953294707&partnerID=8YFLogxK

U2 - 10.1109/INFCOM.2010.5462197

DO - 10.1109/INFCOM.2010.5462197

M3 - Conference contribution

SN - 9781424458363

BT - 2010 Proceedings IEEE INFOCOM

ER -