Is your solution accurate? A fault-oriented performance prediction method for enhancing communication network reliability

Communication network performance prediction is a basic tool for designing robust networks, and its precision directly determines the accuracy of downstream task solutions. Modern communication networks possess adaptability, which recovers from faults by updating routing tables and adjusting flow pa...

Full description

Saved in:

Bibliographic Details
Published in:	Reliability engineering & system safety Vol. 256; p. 110793
Main Authors:	Yang, Fang, Ma, Tao, Shu, Nina, Liu, Chunsheng, Wu, Tao, Chang, Chao
Format:	Journal Article
Language:	English
Published:	Elsevier Ltd 01.04.2025
Subjects:	Dynamic queuing network Network adaptive behavioral dynamics Network fault Network performance prediction Network fault Dynamic queuing network Network adaptive behavioral dynamics Network performance prediction
ISSN:	0951-8320
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Communication network performance prediction is a basic tool for designing robust networks, and its precision directly determines the accuracy of downstream task solutions. Modern communication networks possess adaptability, which recovers from faults by updating routing tables and adjusting flow paths, etc. However, the existing performance prediction methods do not consider the complex adaptive behavior of the network post-fault, leading to significant prediction errors. In this paper, we propose a fault-oriented communication network performance prediction method based on network adaptive behavioral dynamics and dynamic queuing networks (NAB-DQN). This method models the dynamic behavior of the network after faults from the routing perspective and transforms them into M/M/1/K capacity-limited dynamic Jackson open network. Finally, NAB-DQN implements flow-level network performance prediction based on flow path dependency characteristics. We validate the accuracy, generalizability, and efficiency of the algorithm through extensive experiments from different network scenarios (e.g., topology, Flow Density, routing protocol, network size) and fault scenarios (e.g., router, link, port). The experimental results show that NAB-DQN achieves less than 10% error in most scenarios compared to the Discrete Event Simulator (DES) and achieves more than two data orders of magnitude speedup.
ISSN:	0951-8320
DOI:	10.1016/j.ress.2024.110793