Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 25 Apr 1999 09:31:54 -0400 (EDT)
From:      Mikhail Teterin <mi@kot.ne.mediaone.net>
To:        ken@plutotech.com (Kenneth D. Merry)
Cc:        dnelson@emsphone.com, freebsd-scsi@FreeBSD.ORG
Subject:   Re: writing slower with SoftUpdates
Message-ID:  <199904251331.JAA04525@kot.ne.mediaone.net>
In-Reply-To: <199904250636.AAA49618@panzer.plutotech.com> from "Kenneth D. Merry" at "Apr 25, 1999 00:36:51 am"

next in thread | previous in thread | raw e-mail | index | archive | help
Kenneth D. Merry once stated:

mi: WCE was ON on both drives:
mi: Checking SCSI drives for write-cache:
mi: <SEAGATE ST39102LW 0006>           at scbus0 target 8 lun 0 (pass0,da0)
mi: <SEAGATE ST39102LW 0006>           at scbus0 target 9 lun 0 (pass1,da1)

mi: I disabled WCE on both and re-ran the iozone tests. No noticable
mi: changes from the previous numbers... What really bothers me, is the
mi: fact that reading speed speed is consistently only 13-14Mb/s on one
mi: disk, while it is 17-18 (same as writing) on the other. 

=> =	camcontrol defects -u 0 -f phys -G
=> 	Got 0 defects.
=> 	camcontrol defects -u 1 -f phys -G
=> 	Got 0 defects.

=> 	camcontrol defects -u 0 -f phys -PG | wc -l
=> 	Got 868 defects:
=> 	     868
=> 	camcontrol defects -u 1 -f phys -PG | wc -l
=> 	Got 141 defects:
=> 	     141
=> 
=> Khmm, the second disk is indeed faster for writing and a LOT faster
=> for reading.  Does this number of defects justify an exhange request
=> -- we just bought both disks?

=I don't think it's too far out of line for a 9 gig disk, but I suppose that
=could be one possible explanation for the speed difference.

The difference (for reading) is 13-14Mb vs 17-18Mb per second -- kind of
huge for seemingly identical drives.

=What happens when you read off the raw device, instead of from the
=filesystem?  It might be nice to elminate some of the VM system from the
=equation.  Also, is this on -current or -stable or what?

This is on a very recent -stable... I'm going to update it to the very
latest and retry...

=> Why do SoftUpdates slow things down on both disks, anyway? With
=> and without WCE -- the numbers don't change noticeably...

=Well, perhaps there's some sort of VM problem that's causing the soft
=updates slowness?  I really don't know what's going on.

The two 256Mb paging areas I have, are also on the identical locations
on the disks, but they are not even touched having 128Mb of RAM...

=So, that seems to point to maybe some sort of VM-type problem.  Is iozone
=eating a lot of CPU time when you run benchmarks?

Not at all...

=> => The machine is 350MHz PentumII with 128Mb of 100MHz RAM. The disks
=> => are LVD being the only targets on the LVD outlet of Adaptec 2940U2W.

=> =One other thing to look at is where you are on the disk. I/O
=> =performance will be better on the outer diameter tracks.
=> 
=> I know, but the partitions are identical:
=> 
=> 	mi@labservn:~ (95) mount | grep /tmp
=> /dev/da0s1e on /tmp (local, soft-updates, writes: sync 4 async 8014)
=> /dev/da1s1e on /var/tmp (local, soft-updates, writes: sync 23 async 4801)
=> 	mi@labservn:~ (96) ( disklabel da0 ; disklabel da1 ) | grep ' e:'
=>   e:  1048576   655360    4.2BSD        0     0     0   # (Cyl.  963*- 2505*)
=>   e:  1048576   655360    4.2BSD        0     0     0   # (Cyl.  963*- 2505*)

Thanks!

	-mi


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?199904251331.JAA04525>