Re: [tcpm] Increasing the Initial Window - Notes

[Hagen Paul Pfeifer <[e-mail]>]

* rick jones | 2010-11-10 09:20:23 [-0800]:

>Earlier you wrote "New data are always pushed into the network in
>this quantum - a way to much for the standard TCP CC mechanism to tune into a
>steady state for network with a low BDP."  I read that as implying a concern
>about how much data any one connection injects into the network at any moment
>in time.

To clarify this: I especially focus on the start phase of the connection. If
the CC algorithm detects during slow start, congestion avoidance that the
available BDP is larger then the IW it is fine. But if the IW is 10, how
should _new connections_ react, they always will initially transmit 10 packets
- which can be too much - the problem is the initial quantum.

>And I presume, make it configurable.  It is already configurable on a
>per-route basis in the Linux stack.

Yes, and you know the statement of davem and other guys if the discussion goes
in this direction. Davem is aware of the impact of this knob and I am not sure
if everybody is happy about it. I remember a discussion with davem about a
HOT-TO-AGGRESSIVELY-TUNE-YOUR-NETWORK-STACK-HOWTO.xml. If these option are
available they can be misused. I am happy that the per route knobs are a
little bit tricky usable. ;-)

Hagen