FaaStest - Machine learning based cost and performance FaaS optimization

Shay Horovitz, Roei Amos, Ohad Baruch, Tomer Cohen, Tal Oyar, Afik Deri

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

19 Scopus citations

Abstract

With the emergence of Function-as-a-Service (FaaS) in the cloud, pay-per-use pricing models became available along with the traditional fixed price model for VMs and increased the complexity of selecting the optimal platform for a given service. We present FaaStest - an autonomous solution for cost and performance optimization of FaaS services by taking a hybrid approach - learning the behavioral patterns of the service and dynamically selecting the optimal platform. Moreover, we combine a prediction based solution for reducing cold starts of FaaS services. Experiments present a reduction of over 50% in cost and over 90% in response time for FaaS calls.

Original languageEnglish
Title of host publicationEconomics of Grids, Clouds, Systems, and Services - 15th International Conference, GECON 2018, Proceedings
EditorsDaniele D’Agostino, Jörn Altmann, Massimo Coppola, Emanuele Carlini, José Ángel Bañares
PublisherSpringer Verlag
Pages171-186
Number of pages16
ISBN (Print)9783030133412
DOIs
StatePublished - 1 Jan 2019
Externally publishedYes
Event15th International Conference on the Economics of Grids, Clouds, Systems, and Services, GECON 2018 - Pisa, Italy
Duration: 18 Sep 201820 Sep 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11113 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference15th International Conference on the Economics of Grids, Clouds, Systems, and Services, GECON 2018
Country/TerritoryItaly
CityPisa
Period18/09/1820/09/18

Keywords

  • Function as a service
  • Machine learning
  • Serverless

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'FaaStest - Machine learning based cost and performance FaaS optimization'. Together they form a unique fingerprint.

Cite this