*BSD News Article 73006


Return to BSD News archive

#! rnews 3283 bsd
Path: euryale.cc.adfa.oz.au!newshost.anu.edu.au!harbinger.cc.monash.edu.au!nntp.coast.net!zombie.ncsc.mil!news.mathworks.com!newsfeed.internetmci.com!zdc!zdc-e!szdc-e!news
From: "John S. Dyson" <toor@dyson.iquest.net>
Newsgroups: comp.os.linux.networking,comp.unix.bsd.netbsd.misc,comp.unix.bsd.freebsd.misc
Subject: Re: TCP latency
Date: Sat, 06 Jul 1996 21:07:42 -0500
Organization: John S. Dyson's home machine
Lines: 45
Message-ID: <31DF1BEE.41C67EA6@dyson.iquest.net>
References: <4paedl$4bm@engnews2.Eng.Sun.COM> <4qaui4$o5k@fido.asd.sgi.com>
		<4qc60n$d8m@verdi.nethelp.no> <31D2F0C6.167EB0E7@inuxs.att.com>
		<4rfkje$am5@linux.cs.Helsinki.FI> <31DC8EBA.41C67EA6@dyson.iquest.net> <x7ybkxgcx2.fsf@oberon.di.fc.ul.pt>
NNTP-Posting-Host: dyson.iquest.net
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 3.0b5a (X11; I; FreeBSD 2.2-CURRENT i386)
Xref: euryale.cc.adfa.oz.au comp.os.linux.networking:44229 comp.unix.bsd.netbsd.misc:3951 comp.unix.bsd.freebsd.misc:22957

Pedro Roque Marques wrote:
> >     John> be exercised more.
> 
> Well. My belief is that neither Linux neither FreeBSD have thier TCP
> specially designed for high-load (read 1000+ TCP connections). I say
> this because when you enter that zone it starts to make sense having
> special features to deal with a great number of connections in
> TIME-WAIT and SYN-RECV (due to unroutable syn,acks). Also there might
> be alternative solutions to the timer handling problem.
> 
There have been major enhancements added by the FreeBSD team, and
I believe subsequently added to the 4.4Lite/2 stuff that enables
*BSD to perform significantly better than the old *BSD implementations
under load.  There are major scalability issues that apply no matter
what the underlying implementation is like.  (There are certain things
that any implementation has to do.)  There might be more things
that can be done, but it does perform very well under extremely high
TCP loads (with lots of small packets.)  I believe that wcarchive
(with 1000+ FTP sessions, handling mirror scripts, and supporting
alot of small packet sessions), runs at about 30% idle most of the time,
for example.  Of course, it can be made better though, and I am sure
that it will be.

I am sure that the latency issue will be visited sometime, and if the
0-load benchmark is also applicable to a real world situation, it will
be improved.

>     John> One last comment -- if you notice that FreeBSD isn't that
>     John> much slower in the localhost case, it appears that the
> 
> relax. :-) 4.4 BSD networking is still considered to be the leading stuff.
> But i think (in a very biased and personal opinion) that Linux
> networking already beats 4.3 based implementations that are still used
> in lots of comercial Unixes. (humm... the value of this statement is
> really next to 0 since comparing a whole networking implementation
> is kind of silly... take it with a few grains of salt)
> 
I think that what bothered me most is that too much was being made
of the no-load latency figure.  I should have said so a little more
clearly.  Simple benchmarks like the latency benchmark (both
lat_tcp and lat_connect) measure valid results only under the
conditions that they are run.

John