Date: Thu, 07 Mar 2002 10:24:39 -0800 From: "Kevin Oberman" <oberman@es.net> To: Cliff Sarginson <csfbsd@raggedclown.net> Cc: questions@FreeBSD.ORG Subject: Re: Softupdates on root in 4.5-RELEASE Message-ID: <20020307182440.F21725D06@ptavv.es.net> In-Reply-To: Your message of "Thu, 07 Mar 2002 06:33:50 %2B0100." <20020307053350.GA2142@raggedclown.net>
next in thread | previous in thread | raw e-mail | index | archive | help
> Date: Thu, 7 Mar 2002 06:33:50 +0100 > From: Cliff Sarginson <csfbsd@raggedclown.net> > Sender: owner-freebsd-questions@FreeBSD.ORG > > On Wed, Mar 06, 2002 at 10:04:45PM -0500, Brian T. Schellenberger wrote: > > On Wednesday 06 March 2002 09:55 am, Cliff Sarginson wrote: > > | On Wed, Mar 06, 2002 at 09:48:16AM -0500, Brian T. Schellenberger wrote: > > | > On Wednesday 06 March 2002 09:35 am, Cliff Sarginson wrote: > > | > | On Wed, Mar 06, 2002 at 03:19:26AM -0600, Mike Meyer wrote: > > | > | > I personally make the root partitoin larger, enable soft updates, and > > | > | > use SCSI disks with proper tagged queuing support so the write cache > > | > | > isn't a problem. > > | > | > > | > | Can you clarify this a bit. > > | > | With a SCSI disk, tagged queuing etc, soft-updates are not helpful ? > > | > | yes or no ? > > | > > > | > No. That is, they ARE helpful. > > | > > > | > But with tagged queuing on a SCSI disk, write caching isn't dangerous, so > > | > there's no reason not to use it. > > | > > > | > With an IDE disk, write caching IS dangerous. > > | > > > | > Another advantage of SCSI. > > | > > | Ok, so currently this BSD system runs off SCSI with soft-updates on > > | everything, except root. > > | > > | Would you recommend that optimal performance would be achieved, without > > | danger, by enabling write-caching as well ? > > > > Yes, if you have tagged queuing. Check your messages (dmeg). > > > Mounting root from ufs:/dev/da0s2a > da0 at sym0 bus 0 target 6 lun 0 > da0: <QUANTUM ATLAS_V_18_WLS 0230> Fixed Direct Access SCSI-3 device > da0: 80.000MB/s transfers (40.000MHz, offset 31, 16bit), Tagged Queueing > Enabled > > > | And what would you recommend therefore with root, which has neither > > | enabled at the moment ? > > > > Provided it's reasonably large (say, over 100M). > > > Yes it is larger than that. > > > But you can't enable/disable write-caching on a per-partition basis anyway. > Yes, of course. > > > Ok, that sounds good. > Mmm. I am curious and will try this. > One thing for anybody considering wanting to go broke in a hurry, I > personally can verify that SCSI makes a *huge* difference on a system. > Buildworld, compared to doing it on my IDE drives, just blinks by in > comparison. If this may make it even faster I will be impressed :) Make sure that you have options UFS_DIRHASH in your kernel and delete and re-cvsup your entire /usr/src tree. This can make a really big difference in buildworld times. > But the price differential is phenomenal ...at least here in Holland :(. > > e.g 40GB 7200 ATA 100 = 138 Euros (c. $160) > 36.7GB 7200 SCSI U160 = 389 Euros (c. $480) Ouch! I see prices a bit closer, but SCSI is a LOT more $$$. R. Kevin Oberman, Network Engineer Energy Sciences Network (ESnet) Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab) E-mail: oberman@es.net Phone: +1 510 486-8634 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020307182440.F21725D06>