Markov model based disk power management for data intensive workloads

Rajat Garg, Seung Woo Son, Mahmut Kandemir, Padma Raghavan, Ramya Prabhakar

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

19 Citations (Scopus)

Abstract

In order to meet the increasing demands of present and upcoming data-intensive computer applications, there has been a major shift in the disk subsystem, which now consists of more disks with higher storage capacities and higher rotational speeds. These have made the disk subsystem a major consumer of power, making disk power management an important issue. People have considered the option of spinning down the disk during periods of idleness or serving the requests at lower rotational speeds when performance is not an issue. Accurately predicting future disk idle periods is crucial to such schemes. This paper presents a novel disk-idleness prediction mechanism based onMarkov models and explains how this mechanism can be used in conjunction with a three-speed disk. Our experimental evaluation using a diverse set of workloads indicates that (i) prediction accuracies achieved by the proposed scheme are very good (87.5% on average); (ii) it generates significant energy savings over the traditional power-saving method of spinningdown the disk when idle (35.5% on average); (iii) it performs better than a previously proposed multi-speed disk management scheme (19% on average); and (iv) the performance penalty is negligible (less than 1%on average). Overall, our implementation and experimental evaluation using both synthetic disk traces and traces extracted from real applications demonstrate the feasibility of a Markov-model-based approach to saving disk power.

Original languageEnglish (US)
Title of host publication2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009
Pages76-83
Number of pages8
DOIs
StatePublished - Oct 13 2009
Event2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009 - Shanghai, China
Duration: May 18 2009May 21 2009

Publication series

Name2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009

Other

Other2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009
CountryChina
CityShanghai
Period5/18/095/21/09

Fingerprint

Computer applications
Energy conservation
Power management

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Software

Cite this

Garg, R., Son, S. W., Kandemir, M., Raghavan, P., & Prabhakar, R. (2009). Markov model based disk power management for data intensive workloads. In 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009 (pp. 76-83). [5071857] (2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009). https://doi.org/10.1109/CCGRID.2009.67
Garg, Rajat ; Son, Seung Woo ; Kandemir, Mahmut ; Raghavan, Padma ; Prabhakar, Ramya. / Markov model based disk power management for data intensive workloads. 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009. 2009. pp. 76-83 (2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009).
@inproceedings{3f52d94e488b48d6bdb0856308d14b20,
title = "Markov model based disk power management for data intensive workloads",
abstract = "In order to meet the increasing demands of present and upcoming data-intensive computer applications, there has been a major shift in the disk subsystem, which now consists of more disks with higher storage capacities and higher rotational speeds. These have made the disk subsystem a major consumer of power, making disk power management an important issue. People have considered the option of spinning down the disk during periods of idleness or serving the requests at lower rotational speeds when performance is not an issue. Accurately predicting future disk idle periods is crucial to such schemes. This paper presents a novel disk-idleness prediction mechanism based onMarkov models and explains how this mechanism can be used in conjunction with a three-speed disk. Our experimental evaluation using a diverse set of workloads indicates that (i) prediction accuracies achieved by the proposed scheme are very good (87.5{\%} on average); (ii) it generates significant energy savings over the traditional power-saving method of spinningdown the disk when idle (35.5{\%} on average); (iii) it performs better than a previously proposed multi-speed disk management scheme (19{\%} on average); and (iv) the performance penalty is negligible (less than 1{\%}on average). Overall, our implementation and experimental evaluation using both synthetic disk traces and traces extracted from real applications demonstrate the feasibility of a Markov-model-based approach to saving disk power.",
author = "Rajat Garg and Son, {Seung Woo} and Mahmut Kandemir and Padma Raghavan and Ramya Prabhakar",
year = "2009",
month = "10",
day = "13",
doi = "10.1109/CCGRID.2009.67",
language = "English (US)",
isbn = "9780769536224",
series = "2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009",
pages = "76--83",
booktitle = "2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009",

}

Garg, R, Son, SW, Kandemir, M, Raghavan, P & Prabhakar, R 2009, Markov model based disk power management for data intensive workloads. in 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009., 5071857, 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009, pp. 76-83, 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009, Shanghai, China, 5/18/09. https://doi.org/10.1109/CCGRID.2009.67

Markov model based disk power management for data intensive workloads. / Garg, Rajat; Son, Seung Woo; Kandemir, Mahmut; Raghavan, Padma; Prabhakar, Ramya.

2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009. 2009. p. 76-83 5071857 (2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009).

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

TY - GEN

T1 - Markov model based disk power management for data intensive workloads

AU - Garg, Rajat

AU - Son, Seung Woo

AU - Kandemir, Mahmut

AU - Raghavan, Padma

AU - Prabhakar, Ramya

PY - 2009/10/13

Y1 - 2009/10/13

N2 - In order to meet the increasing demands of present and upcoming data-intensive computer applications, there has been a major shift in the disk subsystem, which now consists of more disks with higher storage capacities and higher rotational speeds. These have made the disk subsystem a major consumer of power, making disk power management an important issue. People have considered the option of spinning down the disk during periods of idleness or serving the requests at lower rotational speeds when performance is not an issue. Accurately predicting future disk idle periods is crucial to such schemes. This paper presents a novel disk-idleness prediction mechanism based onMarkov models and explains how this mechanism can be used in conjunction with a three-speed disk. Our experimental evaluation using a diverse set of workloads indicates that (i) prediction accuracies achieved by the proposed scheme are very good (87.5% on average); (ii) it generates significant energy savings over the traditional power-saving method of spinningdown the disk when idle (35.5% on average); (iii) it performs better than a previously proposed multi-speed disk management scheme (19% on average); and (iv) the performance penalty is negligible (less than 1%on average). Overall, our implementation and experimental evaluation using both synthetic disk traces and traces extracted from real applications demonstrate the feasibility of a Markov-model-based approach to saving disk power.

AB - In order to meet the increasing demands of present and upcoming data-intensive computer applications, there has been a major shift in the disk subsystem, which now consists of more disks with higher storage capacities and higher rotational speeds. These have made the disk subsystem a major consumer of power, making disk power management an important issue. People have considered the option of spinning down the disk during periods of idleness or serving the requests at lower rotational speeds when performance is not an issue. Accurately predicting future disk idle periods is crucial to such schemes. This paper presents a novel disk-idleness prediction mechanism based onMarkov models and explains how this mechanism can be used in conjunction with a three-speed disk. Our experimental evaluation using a diverse set of workloads indicates that (i) prediction accuracies achieved by the proposed scheme are very good (87.5% on average); (ii) it generates significant energy savings over the traditional power-saving method of spinningdown the disk when idle (35.5% on average); (iii) it performs better than a previously proposed multi-speed disk management scheme (19% on average); and (iv) the performance penalty is negligible (less than 1%on average). Overall, our implementation and experimental evaluation using both synthetic disk traces and traces extracted from real applications demonstrate the feasibility of a Markov-model-based approach to saving disk power.

UR - http://www.scopus.com/inward/record.url?scp=70349751512&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70349751512&partnerID=8YFLogxK

U2 - 10.1109/CCGRID.2009.67

DO - 10.1109/CCGRID.2009.67

M3 - Conference contribution

AN - SCOPUS:70349751512

SN - 9780769536224

T3 - 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009

SP - 76

EP - 83

BT - 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009

ER -

Garg R, Son SW, Kandemir M, Raghavan P, Prabhakar R. Markov model based disk power management for data intensive workloads. In 2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009. 2009. p. 76-83. 5071857. (2009 9th IEEE/ACM International Symposium on Cluster Computing and the Grid, CCGRID 2009). https://doi.org/10.1109/CCGRID.2009.67