ADSM-L

Re: "True" Tape Capacity With Collocation?

1999-05-17 16:27:07
Subject: Re: "True" Tape Capacity With Collocation?
From: Paul Zarnowski <vkm AT CORNELLC.CIT.CORNELL DOT EDU>
Date: Mon, 17 May 1999 16:27:07 -0400
At 08:21 AM 5/14/99 -0500, Warner, Alan wrote:
>What is the optimal per cent utilization for a tape backup pool, when
>collocation is turned on?

snip

>Does anyone have a good rule of thumb (or any kind of a rule at all) for
>optimal tape utilization with collocation turned on?
>
>Thanks,
>   Alan

Alan,

We have found 50-60% as a good rule of thumb on our smaller tape robots
(60-100 DLT tapes).  With larger robots I am not sure that rule would hold
up.  There is a definite tradeoff to be made in consuming server resources
to increase tape utilizations.  As you crank down the reclaim percentage,
you will expend more server resources moving the data around.

In our environment we find collocation to be essential.  Without it,
restore times would lengthen considerably.  We could mitigate this, I
suppose, by increasing the number of tape storage pools we have and putting
different clients into different storage pools, thus cutting down the max
number of tapes used for any one client, but that would increase the
management costs to keep track of all of the storage pools.

Much of this is dependent on your situation.  Factors that come into play
are the size of the tape robot (slots), the size of the tape media, the
characteristics of the tape and robot as far as how long it takes to mount
media, the number of client nodes you have, and how patient/impatient your
users are when restoring a disk.  For the last point, while it doesn't
happen a whole lot, we find that when a user's disk fails that the user is
not in the best of moods and we want to get their data back to them as
quickly as possible.  The difference between 2 and 4 hours can be the
difference in restoring their data in the same day or the next day.

If your server is constrained (CPU, memory, etc) then cranking down the
reclamation percentage will only make things worse.  The worst thing that
can happen is that you run out of tape storage and server capacity at the
same time.  Counting on an overall higher capacity of your tape robot can
be dangerous if you don't have a big enough server to keep reclaiming the
data as needed.  For our larger tape robots, I have planned on 50% overall
utilization for budgetting purposes.

On a related note, I have thoughts that the reclamation logic in ADSM could
be improved to (1) cut down on the number of tape mounts required, and (2)
attempt to segregate static data from dynamic data, so that tapes
containing static data could achieve a higher average utilization.  I.e.,
if files don't get expired within a certain period of time, they are more
likely to not get expired in the future, so why not segregate them to a
separate tape pool?

..Paul


--
Paul Zarnowski                         Ph: 607-255-4757
Paul Zarnowski                         Ph: 607-255-4757
747 Rhodes Hall, Cornell University    Fx: 607-255-8521
Ithaca, NY 14853-3801                  Em: psz1 AT cornell DOT edu