Towards Autoscaling with Guarantees on Kubernetes Clusters

Stephen Burroughs, Helge Dickel, Martin Van Zijl, Vladimir Podolskiy, Michael Gerndt, Robi Malik, Panos Patros

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

4 Scopus citations

Abstract

Autoscaling is used by cloud providers, microser-vices, and edge computing applications to respond to dynamic load fluctuations. A critical direction of research has focused on providing guarantees under uncertainty that the auto scaling system will work as intended-both at design-time and more importantly, at runtime. In this work, we evaluate the efficacy of three complementary methods: A) deterministic finite automata, B) probabilistic process algebra and C) Proportional-Integral-Derivative (PID) control. We experimentally evaluate their efficacy in modelling and verify autoscaling on clusters managed by Kubernetes. Our results suggest that deterministic modelling can provide theoretically optimal guarantees for small deployments; probabilistic algebras are able to capture stochastic behaviours, but benefit from deterministic templates; and control theory benefits by providing a black-box approach for modelling, verification and control.

Original languageEnglish
Title of host publicationProceedings - 2021 IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion, ACSOS-C 2021
EditorsEsam El-Araby, Vana Kalogeraki, Danilo Pianini, Frederic Lassabe, Barry Porter, Sona Ghahremani, Ingrid Nunes, Mohamed Bakhouya, Sven Tomforde
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages295-296
Number of pages2
ISBN (Electronic)9781665443937
DOIs
StatePublished - 2021
Event2nd IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion, ACSOS-C 2021 - Virtual, Washington, United States
Duration: 27 Sep 20211 Oct 2021

Publication series

NameProceedings - 2021 IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion, ACSOS-C 2021

Conference

Conference2nd IEEE International Conference on Autonomic Computing and Self-Organizing Systems Companion, ACSOS-C 2021
Country/TerritoryUnited States
CityVirtual, Washington
Period27/09/211/10/21

Keywords

  • Autoscaling
  • Cloud Computing
  • Kubernetes
  • Resource Management
  • Self-Adaptive Systems

Fingerprint

Dive into the research topics of 'Towards Autoscaling with Guarantees on Kubernetes Clusters'. Together they form a unique fingerprint.

Cite this