Evolutionary neural network modeling for energy prediction of cloud data centers

Yong Wee Foo, Cindy Goh, Hong Chee Lim, Yun Li

Research output: Contribution to journalArticle

Abstract

Accurate forecasts of data center energy consumptions can help eliminate risks caused by underprovisioning or waste caused by over-provisioning. However, due to nonlinearity and complexity, energy prediction remains a challenge. An added layer of complexity further comes from dynamically changing workloads. There is a lack of physical principle based clear-box models, and existing black-box based methods such neural networks are restrictive. In this paper, we develop an evolutionary neural network as a structurally optimal black-box model to forecast the energy consumption of a dynamic cloud data center. In particular, the approach to evolving an optimal network is developed from several novel mechanisms of a genetic algorithm, such as a structurally-inclusive matrix encoding and species parallelism that help maintain an overall increasing fitness to overcome slow convergence whilst preventing premature dominance. The model is trained using part of the data obtained from a set of MapReduce jobs on a 120-core Hadoop cluster and is then validated against unseen data. The results, both in terms of prediction speed and accuracy, suggest that this evolutionary neural network approach to cloud data center forecast is highly promising.

LanguageEnglish
Article number002
JournalProceedings of Science
Volume15-20-March-2015
DOIs
Publication statusPublished - 4 Mar 2016

Fingerprint

Neural networks
Energy utilization
Genetic algorithms

Keywords

  • energy prediction
  • cloud data centers
  • neural network modeling

Cite this

Foo, Yong Wee ; Goh, Cindy ; Lim, Hong Chee ; Li, Yun. / Evolutionary neural network modeling for energy prediction of cloud data centers. In: Proceedings of Science. 2016 ; Vol. 15-20-March-2015.
@article{180ac008dbad4ea2b4fdc9f02d9606c2,
title = "Evolutionary neural network modeling for energy prediction of cloud data centers",
abstract = "Accurate forecasts of data center energy consumptions can help eliminate risks caused by underprovisioning or waste caused by over-provisioning. However, due to nonlinearity and complexity, energy prediction remains a challenge. An added layer of complexity further comes from dynamically changing workloads. There is a lack of physical principle based clear-box models, and existing black-box based methods such neural networks are restrictive. In this paper, we develop an evolutionary neural network as a structurally optimal black-box model to forecast the energy consumption of a dynamic cloud data center. In particular, the approach to evolving an optimal network is developed from several novel mechanisms of a genetic algorithm, such as a structurally-inclusive matrix encoding and species parallelism that help maintain an overall increasing fitness to overcome slow convergence whilst preventing premature dominance. The model is trained using part of the data obtained from a set of MapReduce jobs on a 120-core Hadoop cluster and is then validated against unseen data. The results, both in terms of prediction speed and accuracy, suggest that this evolutionary neural network approach to cloud data center forecast is highly promising.",
keywords = "energy prediction, cloud data centers, neural network modeling",
author = "Foo, {Yong Wee} and Cindy Goh and Lim, {Hong Chee} and Yun Li",
year = "2016",
month = "3",
day = "4",
doi = "10.22323/1.239.0002",
language = "English",
volume = "15-20-March-2015",
journal = "Proceedings of Science",
issn = "1824-8039",
publisher = "Sissa Medialab Srl",

}

Evolutionary neural network modeling for energy prediction of cloud data centers. / Foo, Yong Wee; Goh, Cindy; Lim, Hong Chee; Li, Yun.

In: Proceedings of Science, Vol. 15-20-March-2015, 002, 04.03.2016.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Evolutionary neural network modeling for energy prediction of cloud data centers

AU - Foo, Yong Wee

AU - Goh, Cindy

AU - Lim, Hong Chee

AU - Li, Yun

PY - 2016/3/4

Y1 - 2016/3/4

N2 - Accurate forecasts of data center energy consumptions can help eliminate risks caused by underprovisioning or waste caused by over-provisioning. However, due to nonlinearity and complexity, energy prediction remains a challenge. An added layer of complexity further comes from dynamically changing workloads. There is a lack of physical principle based clear-box models, and existing black-box based methods such neural networks are restrictive. In this paper, we develop an evolutionary neural network as a structurally optimal black-box model to forecast the energy consumption of a dynamic cloud data center. In particular, the approach to evolving an optimal network is developed from several novel mechanisms of a genetic algorithm, such as a structurally-inclusive matrix encoding and species parallelism that help maintain an overall increasing fitness to overcome slow convergence whilst preventing premature dominance. The model is trained using part of the data obtained from a set of MapReduce jobs on a 120-core Hadoop cluster and is then validated against unseen data. The results, both in terms of prediction speed and accuracy, suggest that this evolutionary neural network approach to cloud data center forecast is highly promising.

AB - Accurate forecasts of data center energy consumptions can help eliminate risks caused by underprovisioning or waste caused by over-provisioning. However, due to nonlinearity and complexity, energy prediction remains a challenge. An added layer of complexity further comes from dynamically changing workloads. There is a lack of physical principle based clear-box models, and existing black-box based methods such neural networks are restrictive. In this paper, we develop an evolutionary neural network as a structurally optimal black-box model to forecast the energy consumption of a dynamic cloud data center. In particular, the approach to evolving an optimal network is developed from several novel mechanisms of a genetic algorithm, such as a structurally-inclusive matrix encoding and species parallelism that help maintain an overall increasing fitness to overcome slow convergence whilst preventing premature dominance. The model is trained using part of the data obtained from a set of MapReduce jobs on a 120-core Hadoop cluster and is then validated against unseen data. The results, both in terms of prediction speed and accuracy, suggest that this evolutionary neural network approach to cloud data center forecast is highly promising.

KW - energy prediction

KW - cloud data centers

KW - neural network modeling

UR - http://www.scopus.com/inward/record.url?scp=84988706928&partnerID=8YFLogxK

U2 - 10.22323/1.239.0002

DO - 10.22323/1.239.0002

M3 - Article

VL - 15-20-March-2015

JO - Proceedings of Science

T2 - Proceedings of Science

JF - Proceedings of Science

SN - 1824-8039

M1 - 002

ER -