[Tmrg] Proposal to increase TCP initial CWND

krasnoj at gmx.at (Stefan Hirschmann) Tue, 20 July 2010 21:20 UTC

From: krasnoj at gmx.at (Stefan Hirschmann)
Date: Tue, 20 Jul 2010 23:20:59 +0200
Subject: [Tmrg] Proposal to increase TCP initial CWND
In-Reply-To: <AANLkTil937lyUzRvUtdqd2qdl9RN7AZ-Mo_cT-dtmqXz@mail.gmail.com>
References: <AANLkTil937lyUzRvUtdqd2qdl9RN7AZ-Mo_cT-dtmqXz@mail.gmail.com>
Message-ID: <4C46133B.2040402@gmx.at>

Lachlan Andrew wrote:
> Greetings TMRG folk,
> 
> [...]

> The basic TMRG-related issue was:  If a data centre changes a TCP
> parameter for *its* flows, how can it estimate the degradation caused
> to *other* flows sharing resources with these flows?  Google is awash
> with data, but it all shares the bias of coming from their data
> centres.  What models are appropriate to infer the impact on the
> unmeasured flows?


Just a thing I want to mention here. It is not only about this draft it 
is more general how we should model Slow-Start:

My own ADSL connection has 1.7 Mb and uses a queue-size of max 110 
packets (last mile router). So I believe the Google results, saying it 
is an improvement for my connection. But what is the message: "We sent 
less than the buffers of the routers can bear and it was faster." Even 
if they would sent the whole 70 KB at once (I think this is the maximum 
size of a Google search result) it would be faster, cause my "last mile" 
router can queue it. If I have a background traffic using most of the 
buffer, there will be lots of packet drops and the transmission will be 
slower.

So the approach has to be: Make sure that there is packet drop else the 
results have no rational conclusion.


Cheers Stefan