Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!ames!sgi!vjs@rhyolite.wpd.sgi.com
From: vjs@rhyolite.wpd.sgi.com (Vernon Schryver)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Evaluating Network Performance
Summary: ttcp
Message-ID: <39287@sgi.SGI.COM>
Date: 1 Aug 89 01:11:12 GMT
References: <8907241815.AA06758@Gauss.sparta.com> <727@dvnspc1.Dev.Unisys.COM>
Sender: daemon@sgi.SGI.COM
Organization: Silicon Graphics, Inc., Mountain View, CA
Lines: 36

In article <727@dvnspc1.Dev.Unisys.COM>, gary@dvnspc1.Dev.Unisys.COM (Gary Barrett) writes:
> 
> I would  be interested in knowing if there are any "standard"
> benchmarks by which to evaluate TCP/IP implementations across a
> variety of different vendor products.
> 
> Gary Barrett
> Unisys 
> Devon Engineering Facility
> Wayne, PA

Sorry to mention this again, but people keep asking.  I promise to shut up
about it for at least 6 weeks.

BRL's Mike Muuss's ttcp is the only reasonable one I know of.  It was not
written at a vendor.  It is flexible about buffer sizes and so forth.  It
does both UDP and TCP.  It is more accurate than some "blast" tests which
tend to measure the remote "inetd" as much as the transport.  It does not
test file system performance, unlike the common use of ftp.  File systems'
performance varies by a factor of >100, and while that is interesting, a
file system is not usually a layer 3 or 4 service.  Ttcp is free.

Ttcp unfortunately does not average several runs.  It also uses sockets
and not TLI.  However, since most current applications use sockets, one
could say that makes ttcp more, not less fair and accurate.  It is a UNIX
creature, and so would require porting to other systems.

If there is a better benchmark by the preceding criteria, I would like
a copy of it.

You can ftp a copy of ttcp.c from sgi.sgi.com or 192.26.63.16 in sgi/src/ttcp.c


Vernon Schryver
Silicon Graphics
vjs@sgi.com