Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 2 Jul 2004 10:10:24 -0400
From:      Bill Moran <wmoran@potentialtech.com>
To:        Charles Swiger <cswiger@mac.com>
Cc:        questions@freebsd.org
Subject:   PMTUD, is it actually usable? (was Re: REPOST: Performance problems with FTP)
Message-ID:  <20040702101024.5a6f9fa2.wmoran@potentialtech.com>
In-Reply-To: <31116F58-CAC1-11D8-9B33-003065ABFD92@mac.com>
References:  <20040629140231.7b57dedf.wmoran@potentialtech.com> <1187B403-C9FC-11D8-99F8-003065ABFD92@mac.com> <20040629153855.300b3888.wmoran@potentialtech.com> <1133CD52-CA07-11D8-99F8-003065ABFD92@mac.com> <20040629230817.0f6d230e.wmoran@potentialtech.com> <31116F58-CAC1-11D8-9B33-003065ABFD92@mac.com>

next in thread | previous in thread | raw e-mail | index | archive | help
If you've been following this thread, you can probably guess the problem.
If not, I'll summarize it briefly.

I'm looking at a performance problem that appears to be caused by a
combination of fragmented packets, and failure of PMTUD to work because
of blocked/dropped ICMP packets.

As I'm trying to track this down further, I'm getting hung up in that
it appears as if so many systems are broken that I can't sort anything
out.

VERY few hosts seem to return ICMP messages when I ping with large packets,
they seem to just silently drop the packet if the nofrag flag is set (either
that or the ICMP messages are getting blocked).

I'm trying this from a number of different locations, so I'm pretty sure
it's not just this ISP.

I wanted to see if anyone else was having a similar experience with PMTUD
and debugging problems with it.  I've been using hping (since the ping in
4.x does not yet support setting the nodefrag flag) to try to identify
what providers are breaking the system, and (so far) I'm finding that
PMTU doesn't work _anywhere_!

Is my testing strategy invalid?  Does anyone have any advice/suggestions/
war stories to aid me in determining how I should handle this?  Should
I just set the MTU on this server to 500 and give up on optimizing MTU
sizes?

-- 
Bill Moran
Potential Technologies
http://www.potentialtech.com



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040702101024.5a6f9fa2.wmoran>