Detection of State Transitions in Network elements: On-box demo

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

Abstract

Modern network devices like routers offer thousands of operational counters. All of them could be important for network monitoring, though their high number makes this process infeasible, often resulting in only a small subset of the counters to be considered for further interpretation and processing. This demo paper showcases the practical use of an unsupervised multivariate online detector, DESTIN [1], which could assist an operator in automatically monitoring all or at least a very large number of counters and exploring inter-dependencies between them to further the operator's understanding of the state of the network. DESTIN can detect changes in the network without any need for predefined KPIs on the router itself.

Original languageEnglish
Title of host publicationProceedings of the IM 2021 - 2021 IFIP/IEEE International Symposium on Integrated Network Management
EditorsToufik Ahmed, Olivier Festor, Yacine Ghamri-Doudane, Joon-Myung Kang, Alberto E. Schaeffer-Filho, Abdelkader Lahmadi, Edmundo Madeira
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages724-725
Number of pages2
ISBN (Electronic)9783903176324
Publication statusPublished - 17 May 2021
Event17th IFIP/IEEE International Symposium on Integrated Network Management, IM 2021 - Virtual, Bordeaux, France
Duration: 17 May 202121 May 2021

Publication series

NameProceedings of the IM 2021 - 2021 IFIP/IEEE International Symposium on Integrated Network Management

Conference

Conference17th IFIP/IEEE International Symposium on Integrated Network Management, IM 2021
Country/TerritoryFrance
CityVirtual, Bordeaux
Period17/05/2121/05/21

Keywords

  • Change detection
  • Machine learning
  • Network management
  • Principal angles

Fingerprint

Dive into the research topics of 'Detection of State Transitions in Network elements: On-box demo'. Together they form a unique fingerprint.

Cite this