Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 27 Aug 2001 18:04:40 -0700 (PDT)
From:      Harkirat Singh <singh@pdx.edu>
To:        Mike Silbersack <silby@silby.com>
Cc:        Dave Zarzycki <zarzycki@FreeBSD.ORG>, Alfred Perlstein <bright@mu.org>, <freebsd-net@FreeBSD.ORG>
Subject:   Re: RFC: SACK/FACK patch port to Current
Message-ID:  <Pine.GSO.4.33.0108271751490.27335-100000@gere.odin.pdx.edu>
In-Reply-To: <Pine.BSF.4.30.0108262259080.72039-100000@niwun.pair.com>

next in thread | previous in thread | raw e-mail | index | archive | help

I agree with your comment that FCAK is only a retransmission algorithm and
many papers recommends that FACK+SACK improves the performance for
long-delay network (for more information look at 1996 SIGCOMM paper).

I would say that it would be nice to have SACK+FACK+NewReno and all have a
sysctl so that user can use it at will. This facility will give the
leverage to people who want to test the performance of TCP in the presence
of NewReno, Sack and Fack.

In my patch FACK will be used in combination with SACK!


Regards,

Harkirat Singh


On Sun, 26 Aug 2001, Mike Silbersack wrote:

>
> On Wed, 22 Aug 2001, singh wrote:
>
> > I even observed that FreeBSD4.3 adopts to NewReno algorith which is a
> > suggestion in RFC-2582 (which talks about NewReno, SACK and FACK), for clients
> > who can not have SACK/FACK, new reno will alleviate the problem of duplicate
> > acks in Fast Recovery stage and partial ack is a better solution as comapre to
> > reno algorithm.
>
> Ok, I looked over the patch more, as well as the RFCs.  Basic SACK support
> seems straightforward according to the RFCs, but FACK is a bit more
> complex.  From what I can tell, FACK isn't a tcp feature as much as a
> retransmission scheme.  This scheme, in turn, has been updated and is now
> called "rate halving".
>
> Is the FACK implementation in this patch the old version, or the
> rate-halving version?  Also, does FACK spill over into non-SACKed
> connections?  I couldn't tell from a quick readthrough.
>
> I've also noticed that while SACK is sysctl disableable, FACK is not.  A
> sysctl for FACK should be added as well so that we can enable/disable it
> at will (as can be done with newreno.)
>
> Thanks,
>
> Mike "Silby" Silbersack
>
>


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.33.0108271751490.27335-100000>