We study the combination of Active Queue Management (AQM) techniques and age-based scheduling disciplines to improve network performance in the Linux packet scheduler. While AQM algorithms like CoDel, FQ-CoDel, and PIE manage queue lengths and ensure fair bandwidth sharing, age-based scheduling (e.g., Least Attained Service, Two-Level Processor Sharing) reduces mean flow completion time by prioritizing short flows. While their individual benefits are known, the integration of these orthogonal features had not been analysed so far. We propose a hybrid queuing solution using standard Linux utilities (tc, nftables) tested with state-of-the art open source traffic generators to generate realistic Pareto-distributed traffic. Our experiments on 1 ~Gb / s and 10 ~Gb / s links demonstrate that implementing age-based scheduling policies on top of AQM strategies significantly reduces flow completion time, with gains ranging from 3% to 20%. These improvements are achieved without increasing CPU load and while preserving or improving fairness among flows. The solution is easily implementable on any recent Linux kernel, and we provide all the code we realized to replicate and improve our results.

Age-Based CoDel and His Friends. Improving the Linux Scheduler with 2-Level AQM Disciplines

Moschini, Giovanni;Marin, Andrea;Maccari, Leonardo
2025

Abstract

We study the combination of Active Queue Management (AQM) techniques and age-based scheduling disciplines to improve network performance in the Linux packet scheduler. While AQM algorithms like CoDel, FQ-CoDel, and PIE manage queue lengths and ensure fair bandwidth sharing, age-based scheduling (e.g., Least Attained Service, Two-Level Processor Sharing) reduces mean flow completion time by prioritizing short flows. While their individual benefits are known, the integration of these orthogonal features had not been analysed so far. We propose a hybrid queuing solution using standard Linux utilities (tc, nftables) tested with state-of-the art open source traffic generators to generate realistic Pareto-distributed traffic. Our experiments on 1 ~Gb / s and 10 ~Gb / s links demonstrate that implementing age-based scheduling policies on top of AQM strategies significantly reduces flow completion time, with gains ranging from 3% to 20%. These improvements are achieved without increasing CPU load and while preserving or improving fairness among flows. The solution is easily implementable on any recent Linux kernel, and we provide all the code we realized to replicate and improve our results.
2025
Proceedings of the 2025 21st International Conference on Network and Service Management: AI and Sustainability in the Future of Network and Service Management, CNSM 2025
File in questo prodotto:
File Dimensione Formato  
main.pdf

accesso aperto

Tipologia: Documento in Pre-print
Licenza: Accesso libero (no vincoli)
Dimensione 358.84 kB
Formato Adobe PDF
358.84 kB Adobe PDF Visualizza/Apri

I documenti in ARCA sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/10278/5116947
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? ND
social impact