Very bad FreeBSD SCSI RAID5 write speed performance

nihaopaul nihaopaul at hotmail.com
Thu Nov 13 04:43:31 PST 2003


what version drives atlas 10k << is that the IV drives? if so what firmware
are you running?

Paul

----- Original Message ----- 
From: "Gary Stanley" <gary at outloud.org>
To: "Martin Blapp" <mb at imp.ch>
Cc: <freebsd-scsi at freebsd.org>
Sent: Thursday, November 13, 2003 7:38 PM
Subject: Re: Very bad FreeBSD SCSI RAID5 write speed performance


> Well.
>
> W/O softupdates
>
> root at 64:[/tmp/blah]>dd if=/dev/zero of=/tmp/blah/blah
> 89472+0 records in
> 89471+0 records out
> 45809152 bytes transferred in 8.546312 secs (5360108 bytes/sec)
>
> W/ Softupdates
> root at 64:[/mnt]>dd if=/dev/zero of=/mnt/blah
> 116160+0 records in
> 116159+0 records out
> 59473408 bytes transferred in 10.080428 secs (5899889 bytes/sec)
>
> Just about the same. But this machine's write caching settings for the
> controller (LSI MegaRaid) are not correct for Raid 5, and the machine's
> under some moderate load, so when I get a chance to correct them I'll
> repost my findings. I always seem to suffer heavy write speed losses w/
> Raid5, but I can't complain because the read speeds make up for it.
>
> Also, I use U320 Maxtor Atlas 10K drives, MegaRaid (Latest Firmware,
broken
> Cache Settings)
>
>
>
>
>
> At 05:33 AM 11/13/2003, you wrote:
>
> >Hi all,
> >
> >We are currently testing some raid configurations with Linux
> >and FreeBSD.
> >
> >It seems that FreeBSD does something wrong. An adaptec 3200S
> >controller and a IBM ServeRaid5i controller are both times slower
> >with RAID5 on FreeBSD as they are on linux.
> >
> >Both controllers are batterie-backed, have cacheing enabled.
> >
> >Does anyone with these controllers (drivers) can take some
> >numbers of the write-performance ?
> >
> >If needed I can provide numbers (taken with iozone). I've tested it
> >on a async mounted filesystem on both FreeBSD and Linux. And it is not
> >only a benchmark problem. We suffer on this problem on production
> >machines. The production machines perform a bit better since they
> >run 4.X, but still bad (Linux is 3,5 times faster), on CURRENT Linux
> >is 4 times faster.
> >
> >Martin
> >
> >Martin Blapp, <mb at imp.ch> <mbr at FreeBSD.org>
> >------------------------------------------------------------------
> >ImproWare AG, UNIXSP & ISP, Zurlindenstrasse 29, 4133 Pratteln, CH
> >Phone: +41 61 826 93 00 Fax: +41 61 826 93 01
> >PGP: <finger -l mbr at freebsd.org>
> >PGP Fingerprint: B434 53FC C87C FE7B 0A18 B84C 8686 EF22 D300 551E
> >------------------------------------------------------------------
> >
> >_______________________________________________
> >freebsd-scsi at freebsd.org mailing list
> >http://lists.freebsd.org/mailman/listinfo/freebsd-scsi
> >To unsubscribe, send any mail to "freebsd-scsi-unsubscribe at freebsd.org"
>
> _______________________________________________
> freebsd-scsi at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-scsi
> To unsubscribe, send any mail to "freebsd-scsi-unsubscribe at freebsd.org"


More information about the freebsd-scsi mailing list