[Iccrg] Question on RFC 2988 - TCP Retransmission timer

[Mark Allman]

> On 19/09/2007, Mark Allman <mallman at icir.org> wrote:
> >
> > First, if the performance is driven by the magnitude of the RTO then
> > that is a more general problem than with the min RTO, I think.  We have
> > devised much better loss recovery than relying on the RTO and so one
> > would hope that RTOs are rare enough to not be driving performance.
> 
> True, but isn't RTO still required for windows less than 4 segments?

Sort of.  If one uses Limited Transmit (RFC3042) then no, not
necessarily.  But, of course, there are always cases whereby the RTO is
going to be required since the only thing we can actually count on is
the passage of time.  (Or, if we can't count on the passage of time then
who gives a damn, right?! :) )

> That is also the case in which resetting the window is least harmful.

Right.

Also, related to this is Early Retransmit which has been talked about
within the IETF on-and-off for quite a number of years and seems like it
is going to be taken on as a TCPM WG work item.  See
draft-allman-tcp-early-rexmt-05.txt.

> Could the optimal tradeoff possibly depend on the size of the window
> somehow, starting fairly agressive, and then plateauing for cwnd > 8
> or something?

I think someone could make a short-timeout sort of scheme for small
windows.  It'd need to be worked out, but it seems plausible to me.

allman



-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 185 bytes
Desc: not available
Url : http://oakham.cs.ucl.ac.uk/pipermail/iccrg/attachments/20070919/917726e3/attachment.bin