Date: Tue, 30 Mar 1999 12:36:37 +0100 (IST) From: Nick Hilliard <nick@iol.ie> To: nick@iol.ie (Nick Hilliard) Cc: grog@lemis.com, tom@sdf.com, nick@iol.ie, freebsd-scsi@FreeBSD.ORG Subject: Re: dpt raid-5 performance Message-ID: <199903301136.MAA20308@beckett.earlsfort.iol.ie> In-Reply-To: <199903211417.OAA28733@beckett.earlsfort.iol.ie> from "Nick Hilliard" at Mar 21, 99 02:17:13 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> Does anyone know if a 2044W card creates exactly the same RAID structure on
> a disk array as a 3334UW? I have both cards lying around the place at the
> moment, and it would trivial to run benchmarks for both systems just by
> replacing the card on the machine.
I've included some results from Bonnie below. Just for reference, the
system is a 200Mhz P5 with 128K RAM running 3.1-RELEASE with softupdates not
enabled. The /W card is a DPT-2044W; the /U card is a 3334UW, and each card
has 64M standard DRAM. The 3334UW was used to create the RAID-5 array in
each instance and the slice size is given in each case.
The random seeks column is done with 10 seekers rather than the standard 3.
It looks like the 2044W is badly cpu-bound. I guess I won't be buying one
of those again. The 3334UW has a 20Mhz 68020, not a 68040 as someone else
said. It also looks like it's cpu-bound while doing raid writes.
Tom wrote:
> Small strip sizes are good for single-user situations (like running
> Bonnie), because the IO load will be split over all drives.
The results below show that this is not the case. Unfortunately, I didn't
bother checking up the DPT performance counters for various reasons.
Perhaps some day.
> Large strip
> sizes are good for multi-user situations, where the extra overhead of the
> transactions becomes a problem.
The system is actually going to be an FTP server which means lots of cached
random reads with very few writes. It's not going to be a hugely busy ftp
server -- the ftp daemon limit is probably going to be set at 150 or 200
processes.
Nick Hilliard
Ireland On-Line System Operations
-------Sequential Output-------- ---Sequential Input-- --Random--
-Per Char- --Block--- -Rewrite-- -Per Char- --Block--- --Seeks---
Machine MB K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU K/sec %CPU /sec %CPU
8k/W 256 520 7.1 507 1.6 376 2.2 1273 18.1 1271 4.6 156.2 6.5
16k/W 256 541 7.3 525 1.6 421 2.5 2634 37.8 2695 9.7 180.0 7.5
32k/W 256 560 7.6 533 1.7 450 2.6 3536 50.4 3560 12.9 196.1 7.6
64k/W 256 571 7.8 540 1.7 474 2.8 3740 53.4 3779 13.4 204.3 8.1
128k/W 256 553 7.6 539 1.5 462 2.3 4101 58.6 4119 15.1 206.4 8.3
256k/W 256 558 7.6 534 1.7 464 2.7 4284 61.3 4282 15.8 206.4 8.1
512k/W 256 541 7.5 491 1.7 458 2.3 4293 59.1 4335 16.2 193.6 6.8
1M/W 256 526 7.2 408 1.3 442 2.6 4414 63.3 4450 16.7 210.5 8.1
8k/U 256 2177 29.7 2078 6.4 1533 9.1 6593 94.9 8377 30.6 262.1 10.7
16k/U 256 2122 29.7 2089 6.7 1585 9.7 6707 96.8 8596 32.7 289.2 12.1
32k/U 256 2162 30.1 2122 6.8 1652 10.2 6726 97.3 8521 31.7 307.2 12.5
64k/U 256 2176 30.3 2137 6.8 1674 10.1 6740 97.6 8788 32.0 310.4 12.9
128k/U 256 2156 29.1 2129 6.8 1675 10.5 6968 96.9 9158 34.2 316.1 13.4
256k/U 256 2149 29.9 2103 6.9 1693 10.8 6664 96.8 8832 33.8 326.8 13.9
512k/U 256 2104 29.0 2039 6.6 1670 10.3 6701 97.2 9468 36.5 322.4 13.5
1M/U 256 1827 25.4 1544 5.0 1442 9.2 6658 96.5 10009 38.7 318.6 13.6
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903301136.MAA20308>