*BSD News Article 72628


Return to BSD News archive

Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!munnari.OZ.AU!news.ecn.uoknor.edu!solace!nntp.uio.no!nntp.uib.no!nntp-bergen.UNINETT.no!nntp-trd.UNINETT.no!not-for-mail
From: sthaug@nethelp.no (Steinar Haug)
Newsgroups: comp.os.linux.networking,comp.unix.bsd.netbsd.misc,comp.unix.freebsd.misc
Subject: Re: TCP latency
Date: 02 Jul 1996 22:42:08 GMT
Organization: Nethelp Consulting, Trondheim, Norway
Lines: 48
Message-ID: <4rc8k0$ec0@verdi.nethelp.no>
References: <4paedl$4bm@engnews2.Eng.Sun.COM> <4pf7f9$bsf@white.twinsun.com>
	<4qad7d$a5l@verdi.nethelp.no> <4qaui4$o5k@fido.asd.sgi.com>
	<4qc60n$d8m@verdi.nethelp.no> <31D2F0C6.167EB0E7@inuxs.att.com>
NNTP-Posting-Host: trane.uninett.no
In-reply-to: "John S. Dyson"'s message of Thu, 27 Jun 1996 15:36:22 -0500
Xref: euryale.cc.adfa.oz.au comp.os.linux.networking:43781 comp.unix.bsd.netbsd.misc:3910

["John S. Dyson"]

|   > Pentium local           250 usec
|   > AMD Linux local         330 usec
|   > AMD FreeBSD local       350 usec
|   > AMD Linux -> Pentium    420 usec
|   > AMD FreeBSD -> Pentium  520 usec
|   > 
|   > So the difference is quite noticeable. Wish I had another P133 here to
|   > test with, but unfortunately I don't.
|   > 
|   All this TCP latency discussion is interesting, but how does this
|   significantly impact performance when streaming data through the
|   connection?  Isn't TCP a streaming protocol?

Yes, *but* a lot of people also use it for (small) transactions, where
latency is important.

|   Was TTCP used in these
|   tests? I would think that if you were doing that many connections per
|   second, TTCP would be better generally (like for WWW servers.)

Of course TTCP would be better, but it's not as readily available, and
therefore standard TCP latency *is* of interest. Yes, I know TTCP is
available in FreeBSD.

|   Isn't this just a connection latency?  Hmmm...  Data
|   througput starts overshadowing connection latency quickly.  Also,
|   there is some latency that does not really imply CPU usage...

The measurements give you some idea of the *minimum* latency. This is
relevant for small, short transactions. With 10 Mbit/s Ethernet, you
can have something like 64 byte of data for a request and 512 byte of
data for the answer before the time to move the data on the wire starts
approaching the 520 usec figure above. With 100 Mbit/s Ethernet, much
more data, of course. (But with 100 Mbit/s Ethernet the figures above
are smaller in the first place.)

|   Interesting data point, but really doesn't appear to impact system
|   performance much if at all.
|   
|   Isn't meaningless benchmarking fun!!!

Again, it all depends on what you use your system for. I certainly agree
that this is only one small factor of the total system performance. For
some people this part is important, for others less so.

Steinar Haug, Nethelp consulting, sthaug@nethelp.no