Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing

Alexandru Uta, Sietse Au, Alexey Ilyushkin, Alexandru Iosup

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

Abstract

Graphs are a natural fit for modeling concepts used in solving diverse problems in science, commerce, engineering, and governance. Responding to the diversity of graph data and algorithms, many parallel and distributed graph-processing systems exist. However, until now these platforms use a static model of deployment: they only run on a pre-defined set of machines. This raises many conceptual and pragmatic issues, including misfit with the highly dynamic nature of graph processing, and could lead to resource waste and high operational costs. In contrast, in this work we explore the benefits and drawbacks of the dynamic model of deployment. Building a three-layer benchmarking framework for assessing elasticity in graph analytics, we conduct an in-depth elasticity study of distributed graph processing. Our framework is composed of state-of-the-art workloads, autoscalers, and metrics, derived from the LDBC Graphalytics benchmark and SPEC RG Cloud Group's elasticity metrics. We uncover the benefits and cost of elasticity in graph processing: while elasticity allows for fine-grained resource management, and does not degrade application performance, we find that graph workloads are sensitive to data migration while leasing or releasing resources. Moreover, we identify non-trivial interactions between scaling policies and graph workloads, which add an extra level of complexity to resource management and scheduling for graph processing.

Original languageEnglish
Title of host publicationProceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018
PublisherInstitute of Electrical and Electronics Engineers, Inc.
Pages381-391
Number of pages11
ISBN (Electronic)9781538683194
DOIs
Publication statusPublished - 1 Nov 2018
Event2018 IEEE International Conference on Cluster Computing, CLUSTER 2018 - Belfast, United Kingdom
Duration: 10 Sep 201813 Sep 2018

Conference

Conference2018 IEEE International Conference on Cluster Computing, CLUSTER 2018
CountryUnited Kingdom
CityBelfast
Period10/09/1813/09/18

Fingerprint

Benchmarking
Elasticity
Processing
Costs
Dynamic models
Scheduling

Keywords

  • Benchmark
  • Dynamic scaling
  • Elasticity
  • Elasticity metrics
  • Graph analytics
  • Graph processing

Cite this

Uta, A., Au, S., Ilyushkin, A., & Iosup, A. (2018). Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing. In Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018 (pp. 381-391). [8514898] Institute of Electrical and Electronics Engineers, Inc.. https://doi.org/10.1109/CLUSTER.2018.00056
Uta, Alexandru ; Au, Sietse ; Ilyushkin, Alexey ; Iosup, Alexandru. / Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing. Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018. Institute of Electrical and Electronics Engineers, Inc., 2018. pp. 381-391
@inproceedings{d0295e405f33454dbf6b0fc668c01fe3,
title = "Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing",
abstract = "Graphs are a natural fit for modeling concepts used in solving diverse problems in science, commerce, engineering, and governance. Responding to the diversity of graph data and algorithms, many parallel and distributed graph-processing systems exist. However, until now these platforms use a static model of deployment: they only run on a pre-defined set of machines. This raises many conceptual and pragmatic issues, including misfit with the highly dynamic nature of graph processing, and could lead to resource waste and high operational costs. In contrast, in this work we explore the benefits and drawbacks of the dynamic model of deployment. Building a three-layer benchmarking framework for assessing elasticity in graph analytics, we conduct an in-depth elasticity study of distributed graph processing. Our framework is composed of state-of-the-art workloads, autoscalers, and metrics, derived from the LDBC Graphalytics benchmark and SPEC RG Cloud Group's elasticity metrics. We uncover the benefits and cost of elasticity in graph processing: while elasticity allows for fine-grained resource management, and does not degrade application performance, we find that graph workloads are sensitive to data migration while leasing or releasing resources. Moreover, we identify non-trivial interactions between scaling policies and graph workloads, which add an extra level of complexity to resource management and scheduling for graph processing.",
keywords = "Benchmark, Dynamic scaling, Elasticity, Elasticity metrics, Graph analytics, Graph processing",
author = "Alexandru Uta and Sietse Au and Alexey Ilyushkin and Alexandru Iosup",
year = "2018",
month = "11",
day = "1",
doi = "10.1109/CLUSTER.2018.00056",
language = "English",
pages = "381--391",
booktitle = "Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018",
publisher = "Institute of Electrical and Electronics Engineers, Inc.",

}

Uta, A, Au, S, Ilyushkin, A & Iosup, A 2018, Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing. in Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018., 8514898, Institute of Electrical and Electronics Engineers, Inc., pp. 381-391, 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018, Belfast, United Kingdom, 10/09/18. https://doi.org/10.1109/CLUSTER.2018.00056

Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing. / Uta, Alexandru; Au, Sietse; Ilyushkin, Alexey; Iosup, Alexandru.

Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018. Institute of Electrical and Electronics Engineers, Inc., 2018. p. 381-391 8514898.

Research output: Chapter in Book / Report / Conference proceedingConference contributionAcademicpeer-review

TY - GEN

T1 - Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing

AU - Uta, Alexandru

AU - Au, Sietse

AU - Ilyushkin, Alexey

AU - Iosup, Alexandru

PY - 2018/11/1

Y1 - 2018/11/1

N2 - Graphs are a natural fit for modeling concepts used in solving diverse problems in science, commerce, engineering, and governance. Responding to the diversity of graph data and algorithms, many parallel and distributed graph-processing systems exist. However, until now these platforms use a static model of deployment: they only run on a pre-defined set of machines. This raises many conceptual and pragmatic issues, including misfit with the highly dynamic nature of graph processing, and could lead to resource waste and high operational costs. In contrast, in this work we explore the benefits and drawbacks of the dynamic model of deployment. Building a three-layer benchmarking framework for assessing elasticity in graph analytics, we conduct an in-depth elasticity study of distributed graph processing. Our framework is composed of state-of-the-art workloads, autoscalers, and metrics, derived from the LDBC Graphalytics benchmark and SPEC RG Cloud Group's elasticity metrics. We uncover the benefits and cost of elasticity in graph processing: while elasticity allows for fine-grained resource management, and does not degrade application performance, we find that graph workloads are sensitive to data migration while leasing or releasing resources. Moreover, we identify non-trivial interactions between scaling policies and graph workloads, which add an extra level of complexity to resource management and scheduling for graph processing.

AB - Graphs are a natural fit for modeling concepts used in solving diverse problems in science, commerce, engineering, and governance. Responding to the diversity of graph data and algorithms, many parallel and distributed graph-processing systems exist. However, until now these platforms use a static model of deployment: they only run on a pre-defined set of machines. This raises many conceptual and pragmatic issues, including misfit with the highly dynamic nature of graph processing, and could lead to resource waste and high operational costs. In contrast, in this work we explore the benefits and drawbacks of the dynamic model of deployment. Building a three-layer benchmarking framework for assessing elasticity in graph analytics, we conduct an in-depth elasticity study of distributed graph processing. Our framework is composed of state-of-the-art workloads, autoscalers, and metrics, derived from the LDBC Graphalytics benchmark and SPEC RG Cloud Group's elasticity metrics. We uncover the benefits and cost of elasticity in graph processing: while elasticity allows for fine-grained resource management, and does not degrade application performance, we find that graph workloads are sensitive to data migration while leasing or releasing resources. Moreover, we identify non-trivial interactions between scaling policies and graph workloads, which add an extra level of complexity to resource management and scheduling for graph processing.

KW - Benchmark

KW - Dynamic scaling

KW - Elasticity

KW - Elasticity metrics

KW - Graph analytics

KW - Graph processing

UR - http://www.scopus.com/inward/record.url?scp=85057285825&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85057285825&partnerID=8YFLogxK

U2 - 10.1109/CLUSTER.2018.00056

DO - 10.1109/CLUSTER.2018.00056

M3 - Conference contribution

SP - 381

EP - 391

BT - Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018

PB - Institute of Electrical and Electronics Engineers, Inc.

ER -

Uta A, Au S, Ilyushkin A, Iosup A. Elasticity in Graph Analytics? A Benchmarking Framework for Elastic Graph Processing. In Proceedings - 2018 IEEE International Conference on Cluster Computing, CLUSTER 2018. Institute of Electrical and Electronics Engineers, Inc. 2018. p. 381-391. 8514898 https://doi.org/10.1109/CLUSTER.2018.00056