Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 6 Nov 2006 21:12:34 +0200
From:      Oles Hnatkevych <don_oles@able.com.ua>
To:        Oliver Fromme <freebsd-geom@freebsd.org>
Cc:        freebsd-perfomance@freebsd.org, pjd@freebsd.org
Subject:   Re[2]: geom stripe perfomance question
Message-ID:  <512766008.20061106211234@able.com.ua>
In-Reply-To: <200611061204.kA6C4FXt079703@lurza.secnetix.de>
References:  <961295086.20061105000919@able.com.ua> <200611061204.kA6C4FXt079703@lurza.secnetix.de>

next in thread | previous in thread | raw e-mail | index | archive | help
Hello, All

It turns out that it was a hardware/controller problem.
The benchmarks with 'dd bs=1m' gave the following results.

Two disks Samsung SP2514N, stripe size 65kb

The controller ICH2 (ATA100), Celeron 800:
ad(x): ~72 Mb/sec
ad1 & ad2 in parallel: ~33 Mb/sec each (disks on separate ATA channels)
stripe: ~65-68 Mb/sec

The controller VIA 82xxxx (ATA133), Athlon 1700+
ad(x): 72-74 Mb/sec
ad1 & ad2 in parallel (same ATA channel): ~26Mb/sec each!
ad1 & ad2 in parallel (different ATA channels): same as one disk!
stripe (different ATA channels): ~144 Mb/sec (even without
kern.geom.stripe.fast option).

What really is surprising that disks on the same ATA channel on VIA
controller behave much worse than half of their full speed. Yet on a
different ATA channels they perform flawlessly.

Seems like the gmirror(8) man page should contain a separate paragraph
telling about peculiarities using old ATA controllers or using disks
on the same ATA channel. ;-)

-- 
 Oles Hnatkevych                          mailto:don_oles@able.com.ua




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?512766008.20061106211234>