Accelerating traffic engineering optimization for segment routing: A recommendation perspective

IF 4.4 2区计算机科学 Q1 COMPUTER SCIENCE, HARDWARE & ARCHITECTURE

Computer Networks Pub Date : 2025-03-22 DOI:10.1016/j.comnet.2025.111224

Linghao Wang, Miao Wang, Chungang Lin, Yujun Zhang

{"title":"Accelerating traffic engineering optimization for segment routing: A recommendation perspective","authors":"Linghao Wang, Miao Wang, Chungang Lin, Yujun Zhang","doi":"10.1016/j.comnet.2025.111224","DOIUrl":null,"url":null,"abstract":"<div><div>Traffic engineering (TE) is important for improving network performance. Recently, segment routing (SR) has gained increasing attention in the TE field. Many segment routing traffic engineering (SR-TE) methods compute optimal routing policies by solving linear programming (LP) problems, which suffer from high computation time. Therefore, various methods have been proposed for accelerating TE optimization. However, prior methods solve individual TE optimization problems from scratch, overlooking valuable information from existing historical solutions. We argue that these data can imply the distribution of optimal solutions for solving future TE problems. In this paper, we provide a new perspective on accelerating SR-TE optimization. First, we generated and analyzed historical solutions of a widely used LP model, and revealed two key findings from the data: Flows are predominantly routed through a small subset of intermediate nodes; similar decisions can be made for some flows. Then, inspired by the findings, we propose RS4SR, the first framework to our knowledge leveraging historical solutions for SR-TE acceleration. It can significantly reduce the size of LP model by performing candidate recommendation and flow clustering. Experiments on real-world topologies and various traffic matrices demonstrate that a simple implementation of RS4SR is sufficient to obtain near-optimal solutions within the time limit of two seconds on large-scale networks, utilizing a small number of historical solutions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"264 ","pages":"Article 111224"},"PeriodicalIF":4.4000,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computer Networks","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1389128625001926","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE","Score":null,"Total":0}

引用次数: 0

Abstract

Traffic engineering (TE) is important for improving network performance. Recently, segment routing (SR) has gained increasing attention in the TE field. Many segment routing traffic engineering (SR-TE) methods compute optimal routing policies by solving linear programming (LP) problems, which suffer from high computation time. Therefore, various methods have been proposed for accelerating TE optimization. However, prior methods solve individual TE optimization problems from scratch, overlooking valuable information from existing historical solutions. We argue that these data can imply the distribution of optimal solutions for solving future TE problems. In this paper, we provide a new perspective on accelerating SR-TE optimization. First, we generated and analyzed historical solutions of a widely used LP model, and revealed two key findings from the data: Flows are predominantly routed through a small subset of intermediate nodes; similar decisions can be made for some flows. Then, inspired by the findings, we propose RS4SR, the first framework to our knowledge leveraging historical solutions for SR-TE acceleration. It can significantly reduce the size of LP model by performing candidate recommendation and flow clustering. Experiments on real-world topologies and various traffic matrices demonstrate that a simple implementation of RS4SR is sufficient to obtain near-optimal solutions within the time limit of two seconds on large-scale networks, utilizing a small number of historical solutions.

查看原文本刊更多论文

求助全文

约1分钟内获得全文求助全文

来源期刊

Computer Networks 工程技术-电信学

CiteScore

10.80

自引率

3.60%

发文量

434

审稿时长

8.6 months

期刊介绍： Computer Networks is an international, archival journal providing a publication vehicle for complete coverage of all topics of interest to those involved in the computer communications networking area. The audience includes researchers, managers and operators of networks as well as designers and implementors. The Editorial Board will consider any material for publication that is of interest to those groups.