ADSM-L

Re: TSM Reclamation and large tape formats

2005-06-09 22:51:21
Subject: Re: TSM Reclamation and large tape formats
From: "Allen S. Rout" <asr AT UFL DOT EDU>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 9 Jun 2005 22:50:52 -0400
==> On Thu, 9 Jun 2005 15:48:21 -0400, "Burton, Robert" <robert.burton AT RBC 
DOT COM> said:

> We are using STK 9940 tape drives and get 300GB - 1 TB of data stored on
> each tape.  I'm trying to explain that with the larger tape sizes it is much
> more difficult to keep the reclamation numbers at the same target levels as
> with the smaller tape volumes.

> We have 2 tsm servers / 50 lan-free clients / approx 2 TB of disk pool and
> we backup 8 TB nightly...

> Between copypool and migration we have limited reclamation cycles...

> Just inquiring on what the rest of the tsm community is experiencing in
> regards to tsm reclamation with large tape...

I tend to view this problem as optimizing tape-hours rather than specifically
improving reclamation.  Richard mentioned some special treatment that can be
useful for large files, I'd amplify his statements for e.g. database work (and
with 8TB coming off 50 clients a night, I'd guess there's at least some DB
work in there).

Something to consider:  If you're doing lan-free for all your clients, then
you're wasting -lots- of tape-minutes while remote boxen do incremental
processing.  Unless you're writing to FILE volumes with e.g. SANergy, you're
losing lots of tape efficiency.


> We are being pushed to improve our reclamation thresholds in order to drive
> down the amount of tape volsers required...


An idea you'll have to get across is that the resource consumption is a
complicated differential-equation surface.   If your managment views tape
count as a sole measure of utility, note that the lower your reclamation
thresholds, the more tape drive heads you'll need.

More, if you completely saturate your tape drives 24x7 with routine work, then
any time any of your units does something unusual storage-wise, they'll
encounter problems getting tape heads.  I have this problem now:  my ERP
organization's database backups often collide with reclamation if they do
things in the middle of the day.

---

Something from which you can take heart: Reclamation is one of those problems
that gets better, not worse.  The volume you didn't get around to reclaiming
when it was 55% reclaimaible will be an easier problem next week when it's
60%, and so on.

This might even help you settle on the reclamation threshold your current
environment can stand: Start with a threshold that will only get the
most-reclaimable volume in your library.  Then nudge the threshold down over
time, until you start noticing that it's taking "too long", reclamation isn't
completing.  Nudge up one notch, and there's what your resources can handle.

At that point you can even measure how many tape-head hours are being consumed
in your reclamation, and use that to make a decent guess about how many more
drives would be necessary to achive a manager-happy threshold.


- Allen S. Rout

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