Department of Mathematics, Computer Science, Control of the ISAE

TCED: TCP Congestion Event Detection

Towards Decorrelating Congestion Detection from TCP

3 November 2009

These last years, the number of TCP variants and minor extensions have greatly increase (see [tcpm], [iccrg] and [tsvwg] IETF mailing lists). Some of them are deeply specialized to specific networks such as wireless LAN or satellites links while others focus on high speed networks. To date, except the historical and generic TCP Newreno/SACK (Selective ACKnowledgment) variant, there exists no universal TCP protocol able to perform indifferently over any kind of networks. The direct observable consequence of these many proposals is that TCP source code is gaining in complexity and that minor extensions proposed, such as for instance F-RTO (Forward Retransmission TimeOut-recovery), do not help in terms of clarification of the source code.

The purpose of this project is to propose a novel transport layer architecture where the congestion event detection algorithm is realized independently of the TCP code. We aim to illustrate the feasibility of this concept by demonstrating that we can either obtain similar performances or also improve the accuracy of this detection outside the TCP stack. The main idea is to determine CE (i.e. the congestion detection) which impact on the TCP flow performance by monitoring the TCP flow itself. The principle is to obtain a detection system, at the edge of a network or at the sender-side which analyses the TCP behaviour through the observation of both data packets and acknowledgments paths.

This project is conjointly realized with:

  • Université de la Réunion
  • Université de Nice

A deeper presentation of this concept is given in this paper and a preliminary code webpage here.

Contact Emmanuel Lochin

Mentions légales | Plan d’accès | Site Map | Follow-up of the site's activity RSS 2.0 |