R: [e2e] TCP Westwood

[Saverio Mascolo]

The probing phase of Reno, i.e. the additive increase part, is still there.

The multiplicative decrease part, i.e. window shrinking after congestion, is
improved because instead of having a "blind" window halving we propose a
window reduction that takes into account the congestion status of the
network.
In other words, when we set cwnd= bandwidth estimate *RTTmin, we set a cwnd
that can be larger or  smaller than cwnd/2, i.e. we set a cwnd that takes
into account network status.

Thus, in case of high congestion, this mechanism can reduce the window more
quickly than Reno.
Moreover, we found that when many TCP connections share a 100Mps bottleneck
capacity Westwood increases fairness  up to 170% respect to Reno.

Saverio

----- Original Message -----
From: Pai-Hsiang Hsiao <shawn at eecs.harvard.edu>
To: Saverio Mascolo <mascolo at poliba.it>
Cc: <end2end-interest at postel.org>
Sent: Saturday, June 23, 2001 6:18 AM
Subject: Re: [e2e] TCP Westwood


> On Fri, 22 Jun 2001, Saverio Mascolo wrote:
>
> > at
http://www-ictserv.poliba.it/mascolo/recent_research/recent_research.htm
>
> No more AIMD? Would not that hurt?
>
> ---
>
> // Shawn
>
>