ADSM-L

Data organization on tape storage pool?

2004-05-19 13:01:01
Subject: Data organization on tape storage pool?
From: Alexander Lazarevich <alazarev AT ITG.UIUC DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Wed, 19 May 2004 12:00:50 -0500
Hi TSMers,

TSM 5.1.6.5 on windows 2K server. Library is overland storage neo 4100
with 60 LTO-2 tapes, and 2 X LTO-2 HP drives. Setup the system last July,
works just fine. But I'm a little curious how data is organized on tape
volumes. I don't have a problem with anything, but I'm very curious what
the logic is to how the server decides what data is put on what volume.

For example, I noticed that data from a single client is on several
different tapes. And as data is put on the tape storage pool, the
server finds the tape which has a volume status of 'filling', and
there is only one such tape, so it dumps data on there regardless of
what client that data belongs to. Also, space reclamations can further
jumble up the data on tapes, so that a client may have data on several
different tapes.

My questions is this: is there any way to control the organization of the
data on the tape storage pool. For example, could I force all data from a
single client to be put on a single tape, and if it needed another tape
for data, it would grab a scratch tape only, and avoid putting data on a
tape that has other client data on it. In addition to that, the server
would probably be forced to have x many tapes with a volume status of
'filling', one for each client that backs up, right? Space reclamations
could work fine, it just deletes the old data, then moves any data from a
tape belonging to that client back onto another tape belonging to that
client.

I suppose in this scenario, there might be a lot of waisted tape space,
because many clients have more data than the 200GB LTO-2 can store, so
each client might have one tape which is only using, in theory, 1% of
capacity.

Anyone have any comments? Should I just leave it alone? Did tivoli study
and come up with the best way to organize the data? Or is this something
that people mess around with to get better performance?

Thanks in advance,

Alex
---                                                               ---
   Alex Lazarevich | Systems Administrator | Imaging Technology Group
    Beckman Institute | University of Illinois | www.itg.uiuc.edu
---                                                               ---

<Prev in Thread] Current Thread [Next in Thread>