ADSM-L

Re: migration - when should it happen?

2002-08-29 14:50:16
Subject: Re: migration - when should it happen?
From: "Mark D. Rodriguez" <mark AT MDRCONSULT DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Thu, 29 Aug 2002 13:35:31 -0500
Coviello, Paul wrote:

HI I have a 2 part question.

       1.  we currently have a controlled migration that happens at 10:00
am everyday.  We have the following comand: update      stgp backuppool
highmig=1 lowmig=0 migproc=3.  We are getting complaints that the data
should stay on the disks the    whole day for restorations. I was thinking
about pushing it out to 4:30.  And then of course we have expiration and the
reclamations and the db backup.  So what should be the order of these?   We
have 30gb for our backuppool.  which    leads to the second question.
       2. What command can I run to get the amount of data that was backed
up last night?  in case we want to set it to 50 or      something.  I just
want to be able to control this so it does not effect the backups at night.

       thanks
       Paul

       Paul J Coviello
       Sr Systems Analyst
       Catholic Medical Center
       2456 Brown Ave
       Manchester, NH 03103
       603 663-5326


Hi Paul,

I see you have gotten a couple of answers already, but I will add my 2
cents worth anyway.  As far as your second question. Petur's select
statement looks quite similar to the one I have written to do the same
thing.  It's you first question that can be quite interesting.  First
let me say that there is no single right answer to this question.  You
must consider all things in your environment and then make a decision.

One of the other response had mentioned turning on "cache=yes" for the
diskpool.  I am not saying that is wrong or that it won't work, however
you should realize that it will but a burden on the system during you
backup window when clients are writing to the diskpool in addition it
will increase the size of you TSM DB and TSM log usage.  What happens is
that when a file is cached the DB keeps tracks 2 objects for the same
file, one in the diskpool and one in the tapepool.  Eventually the
diskpool becomes physically full.  When the next file arrives TSM saves
that file which may cause one or more of the "cached" files to be
dropped, therefore TSM has to to some extra DB transactions to remove
the "cached" file references from the DB.  All of this is consuming cpu
cycles and I/O bandwidth at the time when you need it the most.
Therefore, unless performance is not an issue I do not use the
"cache=yes" option for a diskpool.

So here is what I generally do for my customers and why:

1) I size disk storage pools to hold 2.25 x the amount of nightly data
transferred to the server disk pool.
2) I set "MIGDelay=1" so that I hold data on disk a minimum of 1 day.
3) I do my migrations after business hours, but before my backup window
starts.

The reasoning:
1) Having 2.25 x the normal daily data traffic in my diskpool allows me
to hold 2 days worth of data and have a little head room for the unexpected.
2) The combination of 2 and 3 from above gives me the ability to restore
from disk any thing that has changed in the last 2 days, i.e. all
business day Wednesday my customers will be able to restore from disk
anything that has changed since the Sunday night/Monday morning backup.
3) Another aspect is, since I do my "ba stg diskpool copypool" prior to
migration all of my writes to copy pools are from disk to tape not tape
to tape.  I still do a "ba stg tapepool copypool" just in case something
fell thru the cracks, but this normally runs instantly with no data
having to be moved.

Proper scheduling can really make a difference in how well your TSM
environment runs.  Since scheduling is so important you will also see
people have a variety of opinions on how it should be done.  In fact I
have seen people argue over scheduling like they were religious wars!
The bottom line is you must look at the whole picture not just one
aspect of it.  There are many things to consider besides what is the
backup window for my clients.  You should also consider operator
availability to load/unload tapes from the library, when to tapes
physically get taken offsite, what are the frequency and size of your
restores and many other factors as well.  One last thing to consider,
always make sure that you have about 4 hours a day, preferably during
normal hours, when you have nothing scheduled.  This is to allow you
time to catch up in the event of some unforeseen problem that throws you
off schedule.

Sorry for the lengthy reply but I felt this was an important question
that deserved a detailed answer.  I hope this will help you to make your
decisions in a more informed manner.

--
Regards,
Mark D. Rodriguez
President MDR Consulting, Inc.

===============================================================================
MDR Consulting
The very best in Technical Training and Consulting.
IBM Advanced Business Partner
SAIR Linux and GNU Authorized Center for Education
IBM Certified Advanced Technical Expert, CATE
AIX Support and Performance Tuning, RS6000 SP, TSM/ADSM and Linux
Red Hat Certified Engineer, RHCE
===============================================================================

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