[LINUX:1608] KNFD stress testing.. (fwd)

Yavuz Selim Komur (komur@bilkent.edu.tr)
Mon, 26 Oct 1998 09:40:25 +0200 (EET)


---------- Forwarded message ----------
Date: Fri, 23 Oct 1998 17:31:51 -0700
From: Thomas Davis <tadavis@lbl.gov>
To: linux-kernel <linux-kernel@vger.rutgers.edu>
Subject: KNFD stress testing..

Well, I went and applied the lastest knfs (knfs-981022.tar.gz, from
ftp.kernel.org) on a couple of machines, rebooted, and preceded to beat
up on it. Good news is, it didn't fall down and go boom!

Here's the results:

hpux, v10.xx - amd works, mounts ok.
solaris, v2.6 - amd works, mounts ok, knfs complains about v3 NFS stuff.
Linux 2.0.35 - works fine. autofs mounts/umounts ok.

Now, there really interesting parts, performance loads.

Disclaimer: these was done using time & tar. Time is elapsed time.

Client Server Time DataSize KB/sec
(see below) (min:sec) in KB (df)
----------------------------------------------------------------------
Solaris -> 2.1.125 12:11 4078502 5579

2.1.125 -> 2.1.125 37:27 4078502 1815
(I think the network was busy at this time, and blew these values)

local -> local 12:23 4078502 5480
(same filesystem, copy from one directory to another)

/dev/zero -> zerofile 1:47 2105377 19676
(speed test of system)

2.1.125 -> 2.1.125 5:06 1536273 5020
(client was writing to the server)

2.1.125 -> 2.1.125 4:26 1108929 4168
(client was reading from the server)

Note:

No stress testing was done from the HP's, since they are "retired".

I also have a SGI origin 200 that I can also test against this, but I'd
have to check some stuff first.

The Solaris box was a Sun E450, 4 processors, 1024megabyte of RAM, and 4
Fast Ethernet cards, homed into the same switch (a Cisco 5000).

The Linux servers/clients are:

Intel PII/400Mhz
256 Megabytes of RAM
4 UDMA IDE drives (md0, raid0 striped, combined into a 56gigabyte
volume, ext2fs, created with a 1% reserve factor (which is still too
large))
1 DEC Tulip based nic card, into a full duplex 100baseT switch (Cisco
5500)
knsfd running with 16 processes.

The only time I saw "nfs server not responding" was when I would do a
delete of the directories. And, I'd only see that once in deleting
4gigabytes of data.

The only messages that knfs was generating was at mount time from the
solaris box (complaining about the NFS v3 client stuff, I would guess)

Anyone have ideas on how to speed it up any more? :-)

Also, I'd say someone needs to submit the patches in knfsd to linus.

-- 
------------------------+--------------------------------------------------
Thomas Davis		| PDSF Project Leader
tadavis@lbl.gov		| 
(510) 486-4524		| "Only a petabyte of data this year?"

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/