HP3000-L Archives

January 2000, Week 2

HP3000-L@RAVEN.UTC.EDU

Options: Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Message: [<< First] [< Prev] [Next >] [Last >>]
Topic: [<< First] [< Prev] [Next >] [Last >>]
Author: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Subject:
From:
Denys Beauchemin <[log in to unmask]>
Reply To:
Date:
Fri, 14 Jan 2000 13:49:26 -0600
Content-Type:
text/plain
Parts/Attachments:
text/plain (94 lines)
Your tests are very interesting.  They only show one side of the equation
however.  You should try to do the same test but send the backup to $NULL.
 There are three/four components to a backup.  If the backup is to a local
device, the three components are: the disk subsystem, the software and the
backup device.  If the backup is to a remote device, the fourth component is
the network itself.

Now your tests show how fast the tape drives are able to write the data.  By
backing up to $NULL, you will find out how fast you are able to get the data
off the disk drives and through the system and the software and you will not be
limited by the tape drive.

Once you have that figure, you should be able to see if your devices and your
system/software are in balance.  If you are able to strip the data off the disk
drives at 2-3 MBPS, then the tape devices are the bottleneck.  Since a DDS3
require 2+MBPS to keep it streaming and at top efficiency, I am somewhat
surprised at your achieving only 1.245 MBPS on a DDS-3.  A quick test to $NULL
will be very revealing.  Perhaps you are not able to feed the tape drive fast
enough in your environment.

Kind regards,

Denys. . .

Denys Beauchemin
HICOMP
(800) 323-8863  (281) 288-7438         Fax: (281) 355-6879
denys at hicomp.com                             www.hicomp.com


-----Original Message-----
From:   Simonsen, Larry [SMTP:[log in to unmask]]
Sent:   Friday, 14 January, 2000 12:48 PM
To:     [log in to unmask]
Subject:        testing of backup performance

I know these tests were not conducted with the precision of a testing
laboratory but this is what I have found.  The results were good enough for
me to compare the different options.

On 14 January 2000 Larry Simonsen ran some time trials concerning backups.
The system in use is an hp3000/918LX, on the system channel there were 2
disc and the 2 tape drives.  On a separate channel were 4 disc drives.  All
drives were configured as system volumes.  The 2 tape drives involved in the
testing were a dds-3 (HP1537A) and a dds-2(HP1533A).  The media used were
dds-2 and dds-3 tapes.  The store used the standard MPE/ix store command and
the STATISTICS option to report throughput.  The test involved storing
354,569,040 bytes to tape drives.  The figures below are for throughput
(Kbytes/sec).  System load with other users was <10% during the time of the
testing. As with any testing your results may differ.

Drive compression disabled:
Drive           Media           throughput
Dds-2           dds-2           477
Dds-3           dds-2           476
Dds-3           dds-3           904

Drive compression enabled:
Drive           Media           throughput
Dds-2           dds-2           1158
Dds-3           dds-2           1193
Dds-3           dds-3           1245

Up to this point the files were not isolated to any one drive or channel.  I
moved the files to be store to the channel where the drives were not
connected.

Drive compression enabled:
Drive           Media           throughput
Dds-2           dds-2           1146
Dds-3           dds-2           1268
Dds-3           dds-3           1486

Conclusions:
1.      Drive compression makes a lot if difference to store speed.  The
advantage of keeping the tape continuously moving is not better than the
compression.
2.      Without compression using dds-3 media in dds-3 drives nearly doubles
the throughput.
3.      Having the tape drive on the same channel as data for dds-2
mechanism does not impact performance.
4.      Separating the disc channel and the drive channel seems to make
10-15% difference for a dds-3 drive.


-------------------------------------------------
Larry Simonsen                Phone: 801-489-2450
Flowserve Corporation     Fax: 801-491-1750
PO Box 2200                    http://www.Flowserve.com
Springville, UT 84663      e-mail: [log in to unmask]
-------------------------------------------------
All opinions expressed herein are my own and reflect, in no way, those of my
employer.

ATOM RSS1 RSS2