X-no-Archive:yes
Eric describes a real life experience where intelligent use of robotic devices
helped defer an expensive upgrade.  Thank you Eric, I could not have described
a better case for use of robotic devices.  At least for the write or store
portion.  Let me take this discussion further and explain why the use of serial
mode VS random mode is not very useful, or shall we say, not a complete
solution.

In serial mode, one manually loads a tape from the (small) magazine into the
tape drive.  The system uses the tape when it needs it and when the tape is
finished, either no longer needed by the system or is full, it is ejected and
the next one in the magazine is loaded.  When the end of the magazine is
reached, that's it, everything stops.  If someone wishes to restore something
from one of the tapes in the magazine, an operator must fetch the proper tape
and load it in a drive, somewhere.  There is no automation in the restore
process and there is only a linear automation in the store process.

In random mode, which by the way is the only mode supported by libraries that
have more than one tape drive and more than just a few slots (there are a few
exceptions even to this, I believe some small 2 drive autoloaders can still be
run in sequential mode), the software has full control of the library robotics.
 For the Tuesday backup, it is able to instruct the library to load the tape
from slot 58 into drive 7.  The software keeps track of where the files, or
backups are so when the user wants to restore a file backed up last Tuesday,
the software can say to the library, load tape from slot 58 into drive 2, and
the restore occurs, without ever needing any human intervention.  With a large
enough library, one can keep weeks or months of backups near-line.  With
differential SCSI, libraries can be placed 25 meters away from the system,
sometimes it is possible to place the library in a vault with the tapes.  You
can also get SCSI to Fibre - Fibre to SCSI connections which allow IT
departments to place these libraries off site in fire-proof environments
kilometers away from the systems.  A proper backup product should also be able
to easily make copies, in effect clones, of any backups for off-site storage
purposes.

Driving a robot entails far more than just sending commands to the library, it
encompasses an entire infrastructure that keeps track of which tapes the files
are on .

So, Eric you made brilliant use of sequential mode of a simple autoloader, but
so much more can be achieved with random mode and larger libraries.  Your goal
of always having a tape available for writing was reached.  However, you still
had to label and move the tapes around.  With random mode one can conceivably
load up the library, apportion the slots into various tape pools and then close
the library and forget about it for days, weeks or even months.

If you saved 2.5 hours per night because the system no longer had to wait for
an operator to mount a tape from a pile of scratch tapes right next to the
drive, imagine how much additional time you would save with a library in random
mode during a restore.  In random mode, the library would have the proper tape
loaded and available within a minute, provided the required tape was in the
library.  This is where proper sizing of the library comes in, but you are not
limited by the software now.

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:   Eric H. Sand [SMTP:[log in to unmask]]
Sent:   Thursday, 16 September, 1999 2:12 PM
To:     [log in to unmask]
Subject:        Re: Installing a (C1561B) DDS-1 Autoloader on an HP-3000

<Reply to Denys Beauchemin after Larry>

        <
        Several years ago I worked at large HP3000 shop in the East Bay(Ca)
and I had been given the task of re-organizing the tape libraries across
several HP3000s. We decided to use TAPEPLUS from Unison/Tymlabs and Herstal
auto-loading DAT units. All the tapes were labeled to prevent being
inadvertently written over. Two of the Herstal units were dedicated to
output tapes and loaded with expired labeled DATs. There was (almost) always
a tape ready for output when the system needed it and operator intervention
was not required. As far as being useful Denys, serial mode allowed the
operations center to pick up an extra 2.5 hours of machine time each night
as the batch applications no longer needed to wait for an operator to mount
a tape. This was on a 995/3 way and an expensive upgrade was cancelled.
        >