Packing light

Portable workload performance prediction for the cloud

Jennie Duggan, Yun Chi, Hakan Hacigumus, Shenghuo Zhu, Ugur Cetintemel

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

13 Citations (Scopus)

Abstract

We introduce a new learning-based solution for portable database workload performance prediction. The current state of the art addresses performance prediction for individual, static hardware configurations and thus cannot generalize to new platforms without additional training. In this work, we focus on analytical databases that might be deployed on different hardware configurations, possibly offered by various Infrastructureas-a-Service (IaaS) providers in the cloud. Enabling workload performance predictions that can be ported across hardware configurations and IaaS offerings could significantly help cloud users with their service-purchase decisions and cloud providers with their provisioning decisions. Our solution is based on collaborative filtering modeling and prediction. We applied it to lightweight workload fingerprints that model the characteristics and behavior of concurrent query workloads for carefully selected, abstract hardware configurations. Our preliminary results are derived from experiments with TPC-H and TPC-DS benchmarks on the Amazon and Rackspace clouds. They demonstrate that our techniques can predict analytical workload throughput values for diverse hardware platforms with low training overhead and within approximately 30% of the correct figure.

Original languageEnglish (US)
Title of host publication2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013
Pages258-265
Number of pages8
DOIs
StatePublished - Aug 19 2013
Event2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013 - Brisbane, QLD, Australia
Duration: Apr 8 2013Apr 11 2013

Publication series

NameProceedings - International Conference on Data Engineering
ISSN (Print)1084-4627

Other

Other2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013
CountryAustralia
CityBrisbane, QLD
Period4/8/134/11/13

Fingerprint

Hardware
Collaborative filtering
Throughput
Experiments

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Information Systems

Cite this

Duggan, J., Chi, Y., Hacigumus, H., Zhu, S., & Cetintemel, U. (2013). Packing light: Portable workload performance prediction for the cloud. In 2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013 (pp. 258-265). [6547460] (Proceedings - International Conference on Data Engineering). https://doi.org/10.1109/ICDEW.2013.6547460
Duggan, Jennie ; Chi, Yun ; Hacigumus, Hakan ; Zhu, Shenghuo ; Cetintemel, Ugur. / Packing light : Portable workload performance prediction for the cloud. 2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013. 2013. pp. 258-265 (Proceedings - International Conference on Data Engineering).
@inproceedings{00bf25c8d4314968bebf5836a5390fd2,
title = "Packing light: Portable workload performance prediction for the cloud",
abstract = "We introduce a new learning-based solution for portable database workload performance prediction. The current state of the art addresses performance prediction for individual, static hardware configurations and thus cannot generalize to new platforms without additional training. In this work, we focus on analytical databases that might be deployed on different hardware configurations, possibly offered by various Infrastructureas-a-Service (IaaS) providers in the cloud. Enabling workload performance predictions that can be ported across hardware configurations and IaaS offerings could significantly help cloud users with their service-purchase decisions and cloud providers with their provisioning decisions. Our solution is based on collaborative filtering modeling and prediction. We applied it to lightweight workload fingerprints that model the characteristics and behavior of concurrent query workloads for carefully selected, abstract hardware configurations. Our preliminary results are derived from experiments with TPC-H and TPC-DS benchmarks on the Amazon and Rackspace clouds. They demonstrate that our techniques can predict analytical workload throughput values for diverse hardware platforms with low training overhead and within approximately 30{\%} of the correct figure.",
author = "Jennie Duggan and Yun Chi and Hakan Hacigumus and Shenghuo Zhu and Ugur Cetintemel",
year = "2013",
month = "8",
day = "19",
doi = "10.1109/ICDEW.2013.6547460",
language = "English (US)",
isbn = "9781467353021",
series = "Proceedings - International Conference on Data Engineering",
pages = "258--265",
booktitle = "2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013",

}

Duggan, J, Chi, Y, Hacigumus, H, Zhu, S & Cetintemel, U 2013, Packing light: Portable workload performance prediction for the cloud. in 2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013., 6547460, Proceedings - International Conference on Data Engineering, pp. 258-265, 2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013, Brisbane, QLD, Australia, 4/8/13. https://doi.org/10.1109/ICDEW.2013.6547460

Packing light : Portable workload performance prediction for the cloud. / Duggan, Jennie; Chi, Yun; Hacigumus, Hakan; Zhu, Shenghuo; Cetintemel, Ugur.

2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013. 2013. p. 258-265 6547460 (Proceedings - International Conference on Data Engineering).

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

TY - GEN

T1 - Packing light

T2 - Portable workload performance prediction for the cloud

AU - Duggan, Jennie

AU - Chi, Yun

AU - Hacigumus, Hakan

AU - Zhu, Shenghuo

AU - Cetintemel, Ugur

PY - 2013/8/19

Y1 - 2013/8/19

N2 - We introduce a new learning-based solution for portable database workload performance prediction. The current state of the art addresses performance prediction for individual, static hardware configurations and thus cannot generalize to new platforms without additional training. In this work, we focus on analytical databases that might be deployed on different hardware configurations, possibly offered by various Infrastructureas-a-Service (IaaS) providers in the cloud. Enabling workload performance predictions that can be ported across hardware configurations and IaaS offerings could significantly help cloud users with their service-purchase decisions and cloud providers with their provisioning decisions. Our solution is based on collaborative filtering modeling and prediction. We applied it to lightweight workload fingerprints that model the characteristics and behavior of concurrent query workloads for carefully selected, abstract hardware configurations. Our preliminary results are derived from experiments with TPC-H and TPC-DS benchmarks on the Amazon and Rackspace clouds. They demonstrate that our techniques can predict analytical workload throughput values for diverse hardware platforms with low training overhead and within approximately 30% of the correct figure.

AB - We introduce a new learning-based solution for portable database workload performance prediction. The current state of the art addresses performance prediction for individual, static hardware configurations and thus cannot generalize to new platforms without additional training. In this work, we focus on analytical databases that might be deployed on different hardware configurations, possibly offered by various Infrastructureas-a-Service (IaaS) providers in the cloud. Enabling workload performance predictions that can be ported across hardware configurations and IaaS offerings could significantly help cloud users with their service-purchase decisions and cloud providers with their provisioning decisions. Our solution is based on collaborative filtering modeling and prediction. We applied it to lightweight workload fingerprints that model the characteristics and behavior of concurrent query workloads for carefully selected, abstract hardware configurations. Our preliminary results are derived from experiments with TPC-H and TPC-DS benchmarks on the Amazon and Rackspace clouds. They demonstrate that our techniques can predict analytical workload throughput values for diverse hardware platforms with low training overhead and within approximately 30% of the correct figure.

UR - http://www.scopus.com/inward/record.url?scp=84881416968&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84881416968&partnerID=8YFLogxK

U2 - 10.1109/ICDEW.2013.6547460

DO - 10.1109/ICDEW.2013.6547460

M3 - Conference contribution

SN - 9781467353021

T3 - Proceedings - International Conference on Data Engineering

SP - 258

EP - 265

BT - 2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013

ER -

Duggan J, Chi Y, Hacigumus H, Zhu S, Cetintemel U. Packing light: Portable workload performance prediction for the cloud. In 2013 IEEE 29th International Conference on Data Engineering Workshops, ICDEW 2013. 2013. p. 258-265. 6547460. (Proceedings - International Conference on Data Engineering). https://doi.org/10.1109/ICDEW.2013.6547460