Date: Tue, 17 Sep 1996 12:42:43 +0200 (MET DST) From: grog@lemis.de (Greg Lehey) To: gpalmer@FreeBSD.org (Gary Palmer) Cc: freebsd-hardware@FreeBSD.org (FreeBSD hardware Users), hackers@FreeBSD.org (FreeBSD Hackers) Subject: Re: Slow Etherlink Message-ID: <199609171042.MAA08085@allegro.lemis.de> In-Reply-To: <28202.842917587@orion.webspan.net> from "Gary Palmer" at Sep 16, 96 07:46:27 pm
next in thread | previous in thread | raw e-mail | index | archive | help
Gary Palmer writes:
>
> Jeffrey Barber wrote in message ID
> <01BBA3DF.D5124740@jabpc.rtfm.com>:
>> OK, If we can get past all the sarcasim bull shit!
>
> Perhaps you should phrase your questions better and provide more
> relevant information in that case. Complaining about network speed,
> and then including results which contradict your problem is BOUND to
> get sarcastic responses!
I agree that the original poster gave no useful information and made a
claim that would raise many FreeBSD users' hackles even if it were
proven true, but all this correspondence hasn't done much to identify
whether there is a performance problem. Let's summarize:
1. Ping localhost is a nice way to show the length of some of the
internal paths through the kernel. Jeffrey seemed to confuse it
with Ethernet.
2. The numbers he gave were irrelevant anyway.
3. He claims there are performance problems with telnet and ping.
How about ftp? That's usually the clearest indication of ethernet
throughput.
4. Gary thinks it might be due to the driver. Possibly that depends
on the release, but just by chance I did some tests on Sunday,
after installing Slowaris on my Sparc 2. *Those* results are
interesting.
Here are my results. They weren't really designed to show anything on
FreeBSD, and since I only have one FreeBSD box up and running at the
moment, they're only an indication. I ftp'd a 9 MB file (kernel with
debugging symbols, FWIW) between 3 boxes: a P133 running FreeBSD
2.2-current, a P133 running BSD/OS 2.1, and the SparcStation 2 running
SunOS 4.1.3 and Solaris 2.5. Here the results:
copy to -> /dev/null /tmp/junk
FreeBSD - SunOS 4 1020 kb/s 1020 kb/s
FreeBSD - BSD/OS 1030 kb/s 930 kb/s
FreeBSD - Solaris 2.5 462 kb/s 462 kb/s
I wouldn't complain about the throughput of any of the BSD systems,
though it's interesting how badly BSD/OS fared with a copy to a file.
This was also the only result which varied significantly (between 835
and 960 kb/s) and may be related to the fact that this machine was
also running X and the console on which I did the tests.
The real surprise is Solaris 2.5. The SS2 only has 16 MB of memory,
but all it was doing was receiving the file, so you'd think it could
handle things better than that. Does anybody have any ideas?
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199609171042.MAA08085>