Summary
Asia-Pacific Network Operations and Management Symposium
2022
Session Number:TS1
Session:
Number:TS1-01
Method for Extracting Suspected Faulty Equipment Through Recursive Use of GNN Model
Seiji Sakuma, Ryosuke Sato, Mizuto Nakamura, Kyoko Yamagoe,
pp.-
Publication Date:2022/09/28
Online ISSN:2188-5079
DOI:10.34385/proc.70.TS1-01
PDF download
Summary:
Telecommunications carriers have investigated automating network operations such as failure recovery over the years. When a failure occurs, a large number of alarms (ALMs) are generated from multiple sets of equipment. However, the issuing of ALMs and the number of ALMs change dynamically depending on the situation. Therefore, in order to identify a suspected failure, the scope of the investigation must be specified based on the network topology and the combination of ALMs as a preliminary step. The workflow of such network operations is difficult to define, and efficiently selecting the range to investigate the failure is necessary for rapid failure recovery. However, automating such processes is difficult because in existing search algorithms defining in advance the solution that will terminate a search or conditions under which the search will be completed is impossible. In this paper, we propose a network node search algorithm that uses a graph neural network (GNN) to determine repeatedly the necessity for investigating neighboring equipment to determine the scope of the failure investigation in the network.