Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 19 Oct 2007 10:22:42 -0700
From:      "Kevin Oberman" <oberman@es.net>
To:        Julian Elischer <julian@elischer.org>
Cc:        FreeBSD Net <freebsd-net@freebsd.org>, Maksim Yevmenkin <maksim.yevmenkin@gmail.com>
Subject:   Re: tcp analysis tool? 
Message-ID:  <20071019172242.B02AE45010@ptavv.es.net>
In-Reply-To: Your message of "Thu, 18 Oct 2007 18:31:33 PDT." <471808F5.8010608@elischer.org> 

next in thread | previous in thread | raw e-mail | index | archive | help
--==_Exmh_1192814562_15096P
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline

> Maksim Yevmenkin wrote:
> > On 10/18/07, Julian Elischer <julian@elischer.org> wrote:
> >> does anyone have a favourite tool for analysing tcp flows to easily figure out why a transfer is slow?
> >>
> >> I am hoping for something that can help visualise the flow as one of those
> >> "two timeline poles with lines between them" diagrams..
> >>
> >> (that doesn't require too much extra software to be loaded.
> > 
> > i used tcptrace at one point.
> > 
> > http://jarok.cs.ohiou.edu/software/tcptrace/
> > 
> > works on dumps produced by tcpdump and can do some plots and provides
> > some statistical data.
> > 
> > thanks,
> > max
> 
> thanks..
> 
> I see from the man page it can do what I want!

Both tcptrace and xplot (which is needed to display the graphs) are in
ports. 

tcptrace does the best I have seen of reporting on what is happening
under the covers in a TCP transfer, but it does not exactly tell you why
performance is poor. It does the best job I have seen, but I still find
that I need to dump data into a spread-sheet and generate a plot to get
some data.

Figuring out why systems don't get good TCP performance is a very
difficult thing as there are so many things that can slow it down. We
have lots of GE or 10GE connected sites and only handful have any
clue about tuning for good TCP performance and many have other problems
that make end-system tuning inadequate to fix he problem.

If anyone figures out a better way to tell customers where the problem
is, let alone fixes it, would be a true hero! 
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman@es.net			Phone: +1 510 486-8634
Key fingerprint:059B 2DDF 031C 9BA3 14A4  EADA 927D EBB3 987B 3751

--==_Exmh_1192814562_15096P
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (FreeBSD)
Comment: Exmh version 2.5 06/03/2002

iD8DBQFHGOfikn3rs5h7N1ERAq2gAJ4uwsgXeUUaXkVyvJt9RzVYMmLDcgCcDyPG
wCFUcyKiHtTtf4JMhQh12mY=
=ajzx
-----END PGP SIGNATURE-----

--==_Exmh_1192814562_15096P--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20071019172242.B02AE45010>