Research actions

I. Area:

Congestion control: detection

The problem of path loss rate estimation with TCP protocol has been actively studied by the research community. Several schemes has been proposed in order to accurately estimate the loss rate of a given path. These proposals are used for monitoring or metrology purposed.
However, estimating and identifying the TCP congestion event is currently not addressed.
We define as congestion event: a set of losses occurring on a TCP window which involves a TCP congestion adaptation (i.e. a decrease of the current sending window) [RFC3448]
While nowadays rate-based protocols received a particular attention from the research community, it seems important to better identify these congestion events in order to give an accurate reference to these protocols which aims at behaving in a fair manner with TCP. In this study, we propose a method able to identify a congestion event at the border of an autonomous system. This method is implemented in TCED software:

TCED: TCP Congestion Event Detection

Short flows unfairness

Short lived flows in the Internet suffer from unfairness from long lived flows. For them, a lost is dramatic. Many solutions try to improve this issue according to a end to end approach. But there may be another solution at the level of queue management. Our proposition is known as FavourTail.

Appropriate TCP congestion response

It is known TCP outperforms in the LFN environment. Several solutions tackle this issue. We aims to define a solution according to the delay as a congestion metric in addition to the drop. This work is done under the name TCP Falcon.

RTT and reorder in network

What is the magnitude of theses delays in the Internet ? This action aims to response to this question. With a simple echo request, we try to estimate the density function of theses delays.

II. Ressources:

Projects wiki