Reasoning in agent-based network management

Vilho Räisänen, Kasper Apajalahti

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

136 Downloads (Pure)

Abstract

An increasing complexity of mobile networks and use cases is reflected in requirements for network management. Advanced automation is required to address this challenge in an economically feasible manner. We describe a system based on agent mapping as a platform for automation for network management in 5G networks and beyond. We use classical and probabilistic reasoning for composing solutions to complex requests by means of relatively simple software agents. We describe different variants of the approach in terms of capabilities, ranging from triple store only to system with semantic and probabilistic reasoning functionalities. This approach provides flexibility for network functionality evolution and facilitates software reuse and is compatible with the use of task-specific machine learning algorithms in network management agents. We describe test system used for evaluating the concept, as well as use case evaluation obtained with it.

Original languageEnglish
Title of host publicationIEEE/IFIP Network Operations and Management Symposium
Subtitle of host publicationCognitive Management in a Cyber World, NOMS 2018
PublisherIEEE
Pages1-7
Number of pages7
ISBN (Electronic)9781538634165
DOIs
Publication statusPublished - 6 Jul 2018
MoE publication typeA4 Article in a conference publication
EventIEEE/IFIP Network Operations and Management Symposium - Taipei, Taiwan, Republic of China
Duration: 23 May 201827 May 2018
http://noms2018.ieee-noms.org/

Publication series

NameIEEE/IFIP Network Operations and Management Symposium
PublisherIEEE
ISSN (Electronic)2374-9709

Conference

ConferenceIEEE/IFIP Network Operations and Management Symposium
Abbreviated titleNOMS
CountryTaiwan, Republic of China
CityTaipei
Period23/05/201827/05/2018
Internet address

Fingerprint Dive into the research topics of 'Reasoning in agent-based network management'. Together they form a unique fingerprint.

Cite this