On the latency-accuracy tradeoff in approximate MapReduce jobs

Juan F. Perez, Robert Birke, Lydia Y. Chen

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

3 Citations (Scopus)

Abstract

To ensure the scalability of big data analytics, approximate MapReduce platforms emerge to explicitly trade off accuracy for latency. A key step to determine optimal approximation levels is to capture the latency of big data jobs, which is long deemed challenging due to the complex dependency among data inputs and map/reduce tasks. In this paper, we use matrix analytic methods to derive stochastic models that can predict a wide spectrum of latency metrics, e.g., average, tails, and distributions, for approximate MapReduce jobs that are subject to strategies of input sampling and task dropping. In addition to capturing the dependency among waves of map/reduce tasks, our models incorporate two job scheduling policies, namely, exclusive and overlapping, and two task dropping strategies, namely, early and straggler, enabling us to realistically evaluate the potential performance gains of approximate computing. Our numerical analysis shows that the proposed models can guide big data platforms to determine the optimal approximation strategies and degrees of approximation.

Original languageEnglish (US)
Title of host publicationINFOCOM 2017 - IEEE Conference on Computer Communications
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509053360
DOIs
StatePublished - Oct 2 2017
Event2017 IEEE Conference on Computer Communications, INFOCOM 2017 - Atlanta, United States
Duration: May 1 2017May 4 2017

Conference

Conference2017 IEEE Conference on Computer Communications, INFOCOM 2017
CountryUnited States
CityAtlanta
Period5/1/175/4/17

Fingerprint

Stochastic models
Scalability
Numerical analysis
Scheduling
Sampling
Big data

All Science Journal Classification (ASJC) codes

  • Computer Science(all)
  • Electrical and Electronic Engineering

Cite this

Perez, J. F., Birke, R., & Chen, L. Y. (2017). On the latency-accuracy tradeoff in approximate MapReduce jobs. In INFOCOM 2017 - IEEE Conference on Computer Communications [8057038] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/INFOCOM.2017.8057038
Perez, Juan F. ; Birke, Robert ; Chen, Lydia Y. / On the latency-accuracy tradeoff in approximate MapReduce jobs. INFOCOM 2017 - IEEE Conference on Computer Communications. Institute of Electrical and Electronics Engineers Inc., 2017.
@inproceedings{f847d7d4eb1e42828fd09e4b9d41f906,
title = "On the latency-accuracy tradeoff in approximate MapReduce jobs",
abstract = "To ensure the scalability of big data analytics, approximate MapReduce platforms emerge to explicitly trade off accuracy for latency. A key step to determine optimal approximation levels is to capture the latency of big data jobs, which is long deemed challenging due to the complex dependency among data inputs and map/reduce tasks. In this paper, we use matrix analytic methods to derive stochastic models that can predict a wide spectrum of latency metrics, e.g., average, tails, and distributions, for approximate MapReduce jobs that are subject to strategies of input sampling and task dropping. In addition to capturing the dependency among waves of map/reduce tasks, our models incorporate two job scheduling policies, namely, exclusive and overlapping, and two task dropping strategies, namely, early and straggler, enabling us to realistically evaluate the potential performance gains of approximate computing. Our numerical analysis shows that the proposed models can guide big data platforms to determine the optimal approximation strategies and degrees of approximation.",
author = "Perez, {Juan F.} and Robert Birke and Chen, {Lydia Y.}",
year = "2017",
month = "10",
day = "2",
doi = "10.1109/INFOCOM.2017.8057038",
language = "English (US)",
booktitle = "INFOCOM 2017 - IEEE Conference on Computer Communications",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

Perez, JF, Birke, R & Chen, LY 2017, On the latency-accuracy tradeoff in approximate MapReduce jobs. in INFOCOM 2017 - IEEE Conference on Computer Communications., 8057038, Institute of Electrical and Electronics Engineers Inc., 2017 IEEE Conference on Computer Communications, INFOCOM 2017, Atlanta, United States, 5/1/17. https://doi.org/10.1109/INFOCOM.2017.8057038

On the latency-accuracy tradeoff in approximate MapReduce jobs. / Perez, Juan F.; Birke, Robert; Chen, Lydia Y.

INFOCOM 2017 - IEEE Conference on Computer Communications. Institute of Electrical and Electronics Engineers Inc., 2017. 8057038.

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

TY - GEN

T1 - On the latency-accuracy tradeoff in approximate MapReduce jobs

AU - Perez, Juan F.

AU - Birke, Robert

AU - Chen, Lydia Y.

PY - 2017/10/2

Y1 - 2017/10/2

N2 - To ensure the scalability of big data analytics, approximate MapReduce platforms emerge to explicitly trade off accuracy for latency. A key step to determine optimal approximation levels is to capture the latency of big data jobs, which is long deemed challenging due to the complex dependency among data inputs and map/reduce tasks. In this paper, we use matrix analytic methods to derive stochastic models that can predict a wide spectrum of latency metrics, e.g., average, tails, and distributions, for approximate MapReduce jobs that are subject to strategies of input sampling and task dropping. In addition to capturing the dependency among waves of map/reduce tasks, our models incorporate two job scheduling policies, namely, exclusive and overlapping, and two task dropping strategies, namely, early and straggler, enabling us to realistically evaluate the potential performance gains of approximate computing. Our numerical analysis shows that the proposed models can guide big data platforms to determine the optimal approximation strategies and degrees of approximation.

AB - To ensure the scalability of big data analytics, approximate MapReduce platforms emerge to explicitly trade off accuracy for latency. A key step to determine optimal approximation levels is to capture the latency of big data jobs, which is long deemed challenging due to the complex dependency among data inputs and map/reduce tasks. In this paper, we use matrix analytic methods to derive stochastic models that can predict a wide spectrum of latency metrics, e.g., average, tails, and distributions, for approximate MapReduce jobs that are subject to strategies of input sampling and task dropping. In addition to capturing the dependency among waves of map/reduce tasks, our models incorporate two job scheduling policies, namely, exclusive and overlapping, and two task dropping strategies, namely, early and straggler, enabling us to realistically evaluate the potential performance gains of approximate computing. Our numerical analysis shows that the proposed models can guide big data platforms to determine the optimal approximation strategies and degrees of approximation.

UR - http://www.scopus.com/inward/record.url?scp=85034099178&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85034099178&partnerID=8YFLogxK

U2 - 10.1109/INFOCOM.2017.8057038

DO - 10.1109/INFOCOM.2017.8057038

M3 - Conference contribution

AN - SCOPUS:85034099178

BT - INFOCOM 2017 - IEEE Conference on Computer Communications

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Perez JF, Birke R, Chen LY. On the latency-accuracy tradeoff in approximate MapReduce jobs. In INFOCOM 2017 - IEEE Conference on Computer Communications. Institute of Electrical and Electronics Engineers Inc. 2017. 8057038 https://doi.org/10.1109/INFOCOM.2017.8057038