[Iccrg] SSDT Scope - summary

[Lachlan Andrew]

Greetings Dirceu and Michael,

1. I agree that resuming after "application constrained" is a more
clear-cut case for more rapid increase.  If we can improve that case,
that is already a win.  (BTW, Sally was the one proposing quadrupling.
 My point was that linear increase avoids the excessive overshoot that
comes with exponential increase, be it quadrupling or doubling, while
simultaneously giving a rapid initial start.)

2. It isn't clear to me that the possibility of path changes is a
sufficient reason not to use better slow-start after timeout.
2a) We still start from a low rate and build up, stopping if we get a loss
2b) Most route changes will be between "similar" links -- like
rerouting within a backbone.  We'd only have problems if, for example,
we were forced off a GbE link onto an ADSL backup link
2c) Even if we do overshoot significantly, we'll get a stack of losses
and time out again, but this time with a reduced  ssthresh.

There may be a smarter way to make use of the fact that we have *some*
information about probability distribution of the BDP.

Cheers,
Lachlan

On 30/03/2008, Michael Welzl <michael.welzl at uibk.ac.at> wrote:
> On Fri, 2008-03-28 at 15:10 -0700, Dirceu Cavendish wrote:
>  > Lachlan,
>  >
>  > I've parsed your reference. I guess your focus is on "application
>  > constrained/idle" scenario.
>  > Of course, in this case, there is no ambiguity about RTO causes...
>
>
> ... and the unknown cause is a problem that I have with
>  the idea of, say, quadrupling after an RTO, based on the
>  reasoning that we have an idea of the BDP.
>
>  I think that, after an RTO, the assumption is that we have
>  no idea about anything in the network anymore. FWIW, it could
>  have been caused by a path change
>
>  Cheers,
>
> Michael
>
>
>  PS: Dirceu - I agree about the proposed scope in your previous message
>
>


-- 
Lachlan Andrew  Dept of Computer Science, Caltech
1200 E California Blvd, Mail Code 256-80, Pasadena CA 91125, USA
Ph: +1 (626) 395-8820    Fax: +1 (626) 568-3603
http://netlab.caltech.edu/lachlan