Odd congestion window behaviour [ was: Draft email to freebsd-net ]

[James Healy]

>>I.e. where we keep past connection state and use that
>>as a reference for the initial cwnd. I asked Mark about
>>this in the past.. and he said that his paper was
>>mis-interpreted and this is incorrect behavior. If you
>>have no connections up to a peer you should not use any
>>past value for the cwnd...
> 
> So it's possible that the inital cwnd shouldn't be set by the
hostcache
> at all?
> 
> If this was the case, does that mean we'd just use the rfc 3390 logic
if
> enabled, with fallback to the manual sysctl variables as a last
resort?

I haven't tested it yet, but I just noticed that initial odd behaviour
we observed looks like it's been fixed in tcp_input.c revisions 1.315
(MAIN) / 1.281.2.13 (RELENG_6). It ensures that tp->snd_wnd is
initialised before being used to calculate the initial cwnd.

The follow on question about setting cwnd from the hostcache still
stands I guess, but at least with this fix we won't see transmission
delays while the connection waits for a delayed ack timer to expire.

James

Swinburne University of Technology
CRICOS Provider Code: 00111D

NOTICE
This e-mail and any attachments are confidential and intended only for the use of the addressee. They may contain information that is privileged or protected by copyright. If you are not the intended recipient, any dissemination, distribution, printing, copying or use is strictly prohibited. The University does not warrant that this e-mail and any attachments are secure and there is also a risk that it may be corrupted in transmission. It is your responsibility to check any attachments for viruses or defects before opening them. If you have received this transmission in error, please contact us on +61 3 9214 8000 and delete it immediately from your system. We do not accept liability in connection with computer virus, data corruption, delay, interruption, unauthorised access or unauthorised amendment.

Please consider the environment before printing this email.