How to plan possible effects of reducing versions & how long does it take

[Morgenstern72]

I have to backup backup a 800TB space which results in a backup of 4-15 TB a day.
We use TSM 6.3.4 on linux and TSM Manager 5.3 on windows. Backup is from a linux Frauenhofer Group file system.

Since we constantly battle with a shortage of hardware I would like to know if and how a reduction of versions results in a smaller backup on tape. At the moment we use 30 version to be able to go back one month for files that change every day. Since we run out of tape robot library space I have to plan how much space in tapes woud be freed if I reduce to x versions. Only with these figures I can talk to my boss to get maybe a new extension to my tape robot.

I see no possibility to do this with TSM or TSM Manager and IBM support told me, you can only do it with direct database querys but gave me no support in creating them.

I would be grateful if anyone can help me with a direct example of how this can be done.

===

Second question is: how long does it take that I see a result?
I tried to reduce version to 15 in a test run and can clearly see the version have vanished from the restore client. But no space reduction and no more tapes to reclaim. Can I speed this up somehow?

 

[marclant]

It would be a really complex query to do. In the backups table, you see the backup date, but not the version number. Because expiration works on both age and number of version, so you'd have to somehow count the versions, etc.

And even if you can achieve the above, that just gives you the number of objects, not their size. So you'd have to cross-reference that with another table. So the time to compute that would be very long, hours if not days.

You should see the reduction in space used after expiration and reclamation has run. Depending on your reclamation threshold, it might take a while though. For example, if you use 50% and many volumes are full, but the expired data only brought them to 40% reclaimable, it would still not be eligible for reclamation, even you gained space. You would have to be more aggressive with the reclamation to consolidate the data on as little volumes as possible.

What's your reclamation threshold?

 

[Morgenstern72]

I reclaim at 50%.
Expiration has run and instantly freed 10 tapes, so it has worked

That there is no possibility to check for space took up by versions is quite...disturbing. How can someone plan ahead if you don't know how much space they take?

 

[moon-buddy]

If you start reducing the versions, do expiration, and then reclamation, you see some drop in the number of tape usage. The question is: how significant is the usage drop? This depends on the amount of changed data backed up on a daily basis and the amount of compression TSM applies when it writes to the tape.

Theoretically, as an example, if your data changes 10% daily and you drop the versions from 30 to 15, you should see a 150% reduction in tape usage. How does this translates to GB or TB? Well, you have to know how much data in total has been backed up on a daily basis. This SQL statement may give you a clue:

SELECT CAST(FLOAT(SUM(bytes))/1024/1024/1024 AS DEC(8,2)) AS "Backed up data in GB" FROm summary WHERE activity='BACKUP' AND start_time >{'yyyy-mm-dd 00:00:00'} AND start_time <{'YYYY-MM-DD 00:00:00'}

Substitute the starting and end date in yyyy-mm-dd and YYYY-MM-DD, respectively

With this information, you should be able to determine the amount of tape usage, factoring compression, on a daily basis.