Thread (15 messages) flat view 15 messages, 6 authors, 2011-03-23

Re: [PATCH] tcp: avoid cwnd moderation in undo

From: Yuchung Cheng <hidden>
Date: 2011-03-14 19:10:45

On Mon, Mar 14, 2011 at 3:06 AM, Carsten Wolff [off-list ref] wrote:
The moderation is in place to avoid gigantic segment bursts, which could cause
unnecessary pressure on buffers. In my eyes it's already suboptimal that the
moderation is weakened in the presence of (detected) reordering, let alone
removing it completely.
In the presence of reordering, cwnd is already moderated in Disorder
state before
 entering the (false) recovery.

More importantly, the prior ssthresh is restored and not affected by
moderation. This means, if moderation reduces cwnd to a small value, then cwnd
< ssthresh and TCP will quickly slow-start back to the previous state, without
sending a big burst of segments.

Also, you intended to remove cwnd moderation only from an undo during
recovery, but I think your patch also removes cwnd moderation when the undo is
caused by D-SACK, i.e. most likely after recovery already ended.
Thanks. I will update my patch description. But the same principle
applies that cwnd
should not be moderated on false events. Whether it should be moderated on
reordering or other events is another (complex) design issue. But this
patch does not
touch that.

Yuchung
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help