Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs

Aaron J. Elmore, Divyakant Agrawal, Sudipto Das, Amr El Abbadi, Alexander Pucher, Xifeng Yan

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

25 Citations (Scopus)

Abstract

A multitenant database management system (DBMS) in the cloud must continuously monitor the trade-off between efficient resource sharing among multiple application databases (tenants) and their performance. Considering the scale of hundreds to thousands of tenants in such multitenant DBMSs, manual approaches for continuous monitoring are not tenable. A self-managing controller of a multitenant DBMS faces several challenges. For instance, how to characterize a tenant given its variety of workloads, how to reduce the impact of tenant colocation, and how to detect and mitigate a performance crisis where one or more tenants' desired service level objective (SLO) is not achieved. We present Delphi, a self-managing system controller for a mul-titenant DBMS, and Pythia, a technique to learn behavior through observation and supervision using DBMS-agnostic database level performance measures. Pythia accurately learns tenant behavior even when multiple tenants share a database process, learns good and bad tenant consolidation plans (or packings), and maintains a per-tenant history to detect behavior changes. Delphi detects performance crises, and leverages Pythia to suggests remedial actions using a hill-climbing search algorithm to identify a new tenant placement strategy to mitigate violating SLOs. Our evaluation using a variety of tenant types and workloads shows that Pythia can learn a tenant's behavior with more than 92% accuracy and learn the quality of packings with more than 86% accuracy. During a performance crisis, Delphi is able to reduce 99th percentile latencies by 80%, and can consolidate 45% more tenants than a greedy baseline, which balances tenant load without modeling tenant behavior.

Original languageEnglish
Title of host publicationProceedings of the ACM SIGMOD International Conference on Management of Data
Pages517-528
Number of pages12
DOIs
Publication statusPublished - 29 Jul 2013
Externally publishedYes
Event2013 ACM SIGMOD Conference on Management of Data, SIGMOD 2013 - New York, NY, United States
Duration: 22 Jun 201327 Jun 2013

Other

Other2013 ACM SIGMOD Conference on Management of Data, SIGMOD 2013
CountryUnited States
CityNew York, NY
Period22/6/1327/6/13

Fingerprint

Controllers
Consolidation
History
Monitoring

Keywords

  • Database consolidation
  • Elastic data management
  • Multitenancy
  • Shared nothing architectures
  • Tenant characterization

ASJC Scopus subject areas

  • Software
  • Information Systems

Cite this

Elmore, A. J., Agrawal, D., Das, S., El Abbadi, A., Pucher, A., & Yan, X. (2013). Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs. In Proceedings of the ACM SIGMOD International Conference on Management of Data (pp. 517-528) https://doi.org/10.1145/2463676.2465308

Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs. / Elmore, Aaron J.; Agrawal, Divyakant; Das, Sudipto; El Abbadi, Amr; Pucher, Alexander; Yan, Xifeng.

Proceedings of the ACM SIGMOD International Conference on Management of Data. 2013. p. 517-528.

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

Elmore, AJ, Agrawal, D, Das, S, El Abbadi, A, Pucher, A & Yan, X 2013, Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs. in Proceedings of the ACM SIGMOD International Conference on Management of Data. pp. 517-528, 2013 ACM SIGMOD Conference on Management of Data, SIGMOD 2013, New York, NY, United States, 22/6/13. https://doi.org/10.1145/2463676.2465308
Elmore AJ, Agrawal D, Das S, El Abbadi A, Pucher A, Yan X. Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs. In Proceedings of the ACM SIGMOD International Conference on Management of Data. 2013. p. 517-528 https://doi.org/10.1145/2463676.2465308
Elmore, Aaron J. ; Agrawal, Divyakant ; Das, Sudipto ; El Abbadi, Amr ; Pucher, Alexander ; Yan, Xifeng. / Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs. Proceedings of the ACM SIGMOD International Conference on Management of Data. 2013. pp. 517-528
@inproceedings{4427fe64b03b452388fa51253f153ad3,
title = "Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs",
abstract = "A multitenant database management system (DBMS) in the cloud must continuously monitor the trade-off between efficient resource sharing among multiple application databases (tenants) and their performance. Considering the scale of hundreds to thousands of tenants in such multitenant DBMSs, manual approaches for continuous monitoring are not tenable. A self-managing controller of a multitenant DBMS faces several challenges. For instance, how to characterize a tenant given its variety of workloads, how to reduce the impact of tenant colocation, and how to detect and mitigate a performance crisis where one or more tenants' desired service level objective (SLO) is not achieved. We present Delphi, a self-managing system controller for a mul-titenant DBMS, and Pythia, a technique to learn behavior through observation and supervision using DBMS-agnostic database level performance measures. Pythia accurately learns tenant behavior even when multiple tenants share a database process, learns good and bad tenant consolidation plans (or packings), and maintains a per-tenant history to detect behavior changes. Delphi detects performance crises, and leverages Pythia to suggests remedial actions using a hill-climbing search algorithm to identify a new tenant placement strategy to mitigate violating SLOs. Our evaluation using a variety of tenant types and workloads shows that Pythia can learn a tenant's behavior with more than 92{\%} accuracy and learn the quality of packings with more than 86{\%} accuracy. During a performance crisis, Delphi is able to reduce 99th percentile latencies by 80{\%}, and can consolidate 45{\%} more tenants than a greedy baseline, which balances tenant load without modeling tenant behavior.",
keywords = "Database consolidation, Elastic data management, Multitenancy, Shared nothing architectures, Tenant characterization",
author = "Elmore, {Aaron J.} and Divyakant Agrawal and Sudipto Das and {El Abbadi}, Amr and Alexander Pucher and Xifeng Yan",
year = "2013",
month = "7",
day = "29",
doi = "10.1145/2463676.2465308",
language = "English",
isbn = "9781450320375",
pages = "517--528",
booktitle = "Proceedings of the ACM SIGMOD International Conference on Management of Data",

}

TY - GEN

T1 - Characterizing tenant behavior for placement and crisis mitigation in multitenant DBMSs

AU - Elmore, Aaron J.

AU - Agrawal, Divyakant

AU - Das, Sudipto

AU - El Abbadi, Amr

AU - Pucher, Alexander

AU - Yan, Xifeng

PY - 2013/7/29

Y1 - 2013/7/29

N2 - A multitenant database management system (DBMS) in the cloud must continuously monitor the trade-off between efficient resource sharing among multiple application databases (tenants) and their performance. Considering the scale of hundreds to thousands of tenants in such multitenant DBMSs, manual approaches for continuous monitoring are not tenable. A self-managing controller of a multitenant DBMS faces several challenges. For instance, how to characterize a tenant given its variety of workloads, how to reduce the impact of tenant colocation, and how to detect and mitigate a performance crisis where one or more tenants' desired service level objective (SLO) is not achieved. We present Delphi, a self-managing system controller for a mul-titenant DBMS, and Pythia, a technique to learn behavior through observation and supervision using DBMS-agnostic database level performance measures. Pythia accurately learns tenant behavior even when multiple tenants share a database process, learns good and bad tenant consolidation plans (or packings), and maintains a per-tenant history to detect behavior changes. Delphi detects performance crises, and leverages Pythia to suggests remedial actions using a hill-climbing search algorithm to identify a new tenant placement strategy to mitigate violating SLOs. Our evaluation using a variety of tenant types and workloads shows that Pythia can learn a tenant's behavior with more than 92% accuracy and learn the quality of packings with more than 86% accuracy. During a performance crisis, Delphi is able to reduce 99th percentile latencies by 80%, and can consolidate 45% more tenants than a greedy baseline, which balances tenant load without modeling tenant behavior.

AB - A multitenant database management system (DBMS) in the cloud must continuously monitor the trade-off between efficient resource sharing among multiple application databases (tenants) and their performance. Considering the scale of hundreds to thousands of tenants in such multitenant DBMSs, manual approaches for continuous monitoring are not tenable. A self-managing controller of a multitenant DBMS faces several challenges. For instance, how to characterize a tenant given its variety of workloads, how to reduce the impact of tenant colocation, and how to detect and mitigate a performance crisis where one or more tenants' desired service level objective (SLO) is not achieved. We present Delphi, a self-managing system controller for a mul-titenant DBMS, and Pythia, a technique to learn behavior through observation and supervision using DBMS-agnostic database level performance measures. Pythia accurately learns tenant behavior even when multiple tenants share a database process, learns good and bad tenant consolidation plans (or packings), and maintains a per-tenant history to detect behavior changes. Delphi detects performance crises, and leverages Pythia to suggests remedial actions using a hill-climbing search algorithm to identify a new tenant placement strategy to mitigate violating SLOs. Our evaluation using a variety of tenant types and workloads shows that Pythia can learn a tenant's behavior with more than 92% accuracy and learn the quality of packings with more than 86% accuracy. During a performance crisis, Delphi is able to reduce 99th percentile latencies by 80%, and can consolidate 45% more tenants than a greedy baseline, which balances tenant load without modeling tenant behavior.

KW - Database consolidation

KW - Elastic data management

KW - Multitenancy

KW - Shared nothing architectures

KW - Tenant characterization

UR - http://www.scopus.com/inward/record.url?scp=84880535219&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84880535219&partnerID=8YFLogxK

U2 - 10.1145/2463676.2465308

DO - 10.1145/2463676.2465308

M3 - Conference contribution

SN - 9781450320375

SP - 517

EP - 528

BT - Proceedings of the ACM SIGMOD International Conference on Management of Data

ER -