Ian suggests...
------------------
With regard to DLT, if you work on the basis that time saved on backup
gives you more time for processing etc is it REALLY a more expensive
option?

The short term view will be that it IS expensive but taken with the long
term view it will probably be a cheaper solution...... and just so much
more efficient (providing machine config is right!)
-----------------

The DLT media costs ten times the price of DDS-3 media - but probably only 5
times DDS-4.
The DLT-8000 drives were also about 8 times the DDS-3 drives, but probably
only 4 times DDS-4.

Here are some figures clipped from full backups at a client:-

Using two DDS-3 drives with Turbo Store parallel store
Both on SE SCSI

   MEDIA  1    7:09 PM           11:40 PM           1627 K/S         0
   MEDIA  1    7:09 PM           11:54 PM           1614 K/S         0

Using two DLT-8 drives with Turbo Store parallel store
Each on it's own FWSCSI

   MEDIA  1    7:09 PM           10:22 PM           2516 K/S         0
   MEDIA  1    7:09 PM           10:32 PM           2684 K/S         0

Using two DLT-8 drives with Turbo Store separate jobs, but running together
Each on it's own FWSCSI

 MEDIA  1      7:05 PM            9:08 PM           4616 K/S         0
 MEDIA  1      7:06 PM            9:37 PM           3566 K/S         0


This last set differs from the middle set as follows:-
It was decided to try to load balance the backup as follows;-

The data was moved around between the Autoraid (for Live data) and the JBOD
configured as two private volume sets for other non critical data.
The fullback was split into two separate turbo stores, one for system volume
set, and one for private volume sets.
Both turbo store jobs run together.

The 4616K/S is the Private Volume dataset throughput
The 3566K/S is the Autoraid throughput.

It sort of makes sense because the Autoraid is FWSCSI to the HP3000, but
SESCSI internally, so its throughput is limited by this - maybe if I
specified ;inter on the turbo store command it would help.

The JBOD in the private volumes are eight drives on a single FWSCSI, and
they currently deliver data faster than the Autoraid.


If the client had added two more DDS3 drives and done a 4 ways parallel
store, throughput would probably have exactly doubled, yielding an aggregate
throughput of  6400K/S

If my thinking is straight, if he replaced the two DDS# with 4 DDS4
(assuming they were supported), throughput would have been 50% up on this at
9600K/S. And there would have been backward compatibility, and cheaper
Media, although more points of failure, twice the media and so on.

Fred suggests that we try the Autoraid on it's own, first with ;inter and
the without, and also try the JBOD Private Volumes on their own, and the
with ;inter and see.

Oh, for the luxury of time to do all this.....and we need a quantum leap
forward in backup technology.

But I'll keep the list posted on progress.

Regards


Neil