ADSM-L

using two drives

2000-01-26 10:53:40
Subject: using two drives
From: Thomas Denier <Thomas.Denier AT MAIL.TJU DOT EDU>
Date: Wed, 26 Jan 2000 10:53:40 -0500
> I am using IBM Magstar 3575 L60 library with one drive, but the amount of
data
> to backup is too large and the time too, I can reduce the time of backup
using
> an adicional drive?
> If i have two drives, i can backup using two drives at the same time and
reduce
> in 50% the time of backup?
> I can run two diferent tasks at same time, backup for drive 1 and archive o
> restore for drive 2.

We do two concurrent backups to two tape drives on some of our client systems.
Tests done when this was set up showed almost a factor of two improvement in
overall throughput. As far as we can tell, most of the improvement comes from
having two processes reading from different disks on the client concurrently,
not from doubling the bandwidth available for sending data to tape drives.
Depending on the factor limiting your performance, you might or might not see
a similar improvement. For example, if network capacity is your limiting
factor you probably will not gain anything by having two concurrent backup
streams contending for the available capacity. The client systems that we set
up for multiple streams spent most of their time backing up small numbers of
very large files. We would probably not have gotten such good results if the
clients were backing up large numbers of small files, because of the increased
contention for update access to the ADSM database. Our earliest
attempts at multiple stream backups produced no performance improvement at
all because all the streams for a given client wanted the same output tape.
We ended up with one stream moving data while the rest waited for the tape
to become available. In order to fix this we updated our include/exclude list
to assign the big files mentioned earlier a management class whose destination
storage pool was collocated at the file system level.
<Prev in Thread] Current Thread [Next in Thread>