Date: Tue, 15 Apr 1997 18:43:36 -0500 (CDT) From: Jim Bryant <jbryant@argus> To: kallio@cc.jyu.fi Cc: freebsd-hackers@freebsd.org Subject: Re: Is someone running heavy loaded inn server? Message-ID: <199704152343.SAA12274@argus> In-Reply-To: <Pine.LNX.3.95.970415170634.28021G-100000@itu.cc.jyu.fi> from "Seppo Kallio" at Apr 15, 97 05:14:25 pm
index | next in thread | previous in thread | raw e-mail
In reply:
> Please reply if you are running a heavy loaded News server.
>
> We have Pentium Pro 200 + 128M RAM + 3*4G IBM Wide + 2940 + 3COM 590
>
> + Linux RedHAt 4.1 !!!
^^^^^^^^^^^^^^^^^^^^^ source of problem...
> running inn in heavy load and it does not work (kernel 2.0.30 not at all,
> kernel 2.0.27 maybe) System is slowing down and at last it hangs.
How many instances?
innd is a vm hog if you have any customers at all...
Linux VM leaves a lot to be desired... if you want, do a side by side
comparison with FreeBSD using the NASA/AMES pitest with MX set to a
level just one digit above the largest MX value that will fit in
core... pitest is an excellant CPU/FPU/RAM/VM subsystem
check/benchmark... you can monitor RAM cram with either ps(1) (using
a custom -o string) or top to determine the proper setting of MX...
monitor the run using `/usr/bin/time -l pitest` and compare completion
stats... FreeBSD will win...
for comparison: (MX=23, scalar): 12.6 million digits on a 1.5 Gig sun
supersparc: just under 107 hours, but using only (if I remember right)
just under 700Megs of RAM, g77 + gcc, -O3 -fomit-frame-pointer, on a
loaded development system, with some paging, Solaris 2.4, reniced -20,
4 CPU system, one dedicated to pitest, this is how I burn-in tested a
1G RAM upgrade for reliability...
pitest should run faster on a 128M system [less core, less precision]...
pitest should still be on wcarchive as:
ftp://ftp.cdrom.com/pub/FreeBSD/incoming/pitest.tar.gz
pitest is the ultimate torture test benchmark...
> What kind of experience you have about FreeBSD + inn, what is the version
> of inn + FreeBSD you use?
>
> Does it work?
Yes FreeBSD rox with innd... I think I have run one of [if not the]
highest innd load levels on a FreeBSD box... This was using versions
starting with the SNAPS after the 2.0-RELEASE and up to 2.1-RELEASE...
2.0-RELEASE had broken VM...
the innd version was 1.4sec [as i remember, that ain't too far off
from the actual meggage per instance (1.2-1.4M)]
> What about ccd? Linux mcd seems not to work, or maybe Linux 2940 driver
> has some problems (is it same as in FreeBSD).
ccd seems to still be either alpha-quality or beta-quality from what
the web page says, other than that, i have no idea...
is rodney grimes still in -hackers? he was getting some amazing
striped throughput a couple of years ago...
jim
--
All opinions expressed are mine, if you | "I will not be pushed, stamped,
think otherwise, then go jump into turbid | briefed, debriefed, indexed, or
radioactive waters and yell WAHOO !!! | numbered!" - #1, "The Prisoner"
jbryant@tfs.net - KC5VDJ 2M, 70cm, KPC-3+ - kc5vdj@wv0t.#neks.ks.usa.noam
help
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199704152343.SAA12274>