ADSM-L

Re: DIRMC - Are copypool reclamation performance issues resolved or not.

2005-03-19 05:56:09
Subject: Re: DIRMC - Are copypool reclamation performance issues resolved or not.
From: Paul Fielding <paul AT FIELDING DOT CA>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Sat, 19 Mar 2005 01:06:05 -0700
I'd be interested in more discussion on this point.   My original
understanding was actually a bit different that that.  The impression I had
was that originally directory tree structures were restored before any files
happened, period. Following that, files would be restored.  Net result -
tapes might get mounted twice.

Is my understanding incorrect? (could well be).  If this behavior has indeed
been fixed so that directories are restored as they are hit on the tape
(with a pre-created non-ACLed directory being created first) then it would
indeed make sense that a DIRMC pool is no longer needed.

Is there any documentation on this somewhere I can reference?

regards,

Paul

----- Original Message -----
From: "TSM_User" <tsm_user AT YAHOO DOT COM>
To: <ADSM-L AT VM.MARIST DOT EDU>
Sent: Thursday, March 17, 2005 3:54 PM
Subject: Re: [ADSM-L] DIRMC - Are copypool reclamation performance issues
resolved or not.


If V5.3 in fact only writes in larger blocks in the smaller directories
may take up more space that required.

Still, that issue aside you should no longer need to have a DIRMC pool. At
one time there was a feature (or call it a bug) where every directory had
to be restored as it came up which would cause many more mounts of tape
drives.  For some time now a restore create a directory (without ACL's) so
that the restore can continue. Then when the directory itself is hit it
will simply restore over top of the directory that was created.  This will
ensure each tape is still only ready once.  True, directories are like
small files and just like small files restoring from disk would be faster
but the bug that used to exist has long since been fixed.

Further as people implement file device class storage pools and other disk
only solutions like VTL's I don't see the need for seperating the
directories into a seperate pool.

Kyle

"Rushforth, Tim" <TRushforth AT WINNIPEG DOT CA> wrote:
What in 5.3 warrants new consideration?

The reason we implemented DIRMC is so that when a user restores a file(s)
there are not extra tape mounts to restore the directories We ran into
this on multiple occasions, even when all files were on disk, tape mounts
would occur because the directories were on tape.

Thanks,

Tim Rusforth
City of Winnipeg

-----Original Message-----
From: TSM_User [mailto:tsm_user AT YAHOO DOT COM]
Sent: Wed 3/16/2005 6:48 PM
To: ADSM-L AT VM.MARIST DOT EDU
Cc:
Subject: Re: DIRMC - Are copypool reclamation performance issues resolved
or not.



It is fixed but the reason there have been suggestions to use a file type
device class is because disk pools unline sequential pools are scanned
from begining to end for every storage pool backup. I have had some
customers that have millions of directories in their DIRMC pool. Even when
none change they backup runs from hours on that pool. With a file type
device class only the new volumes would be backed up resulting in a much
faster backup. Now all that being said this new feature in V5.3 warrents
new consideration. My new consideration is to stop using DIRMC pools as
the reason they were created in the first place has also long been fixed.

Kyle

"Thorneycroft, Doug"
wrote:
OK, after spending a large portion of my day reviewing adsm-l post going
back to
2000, I'm still not sure. Does anyone know if there is still a performance
problem
running reclamation on a DIRMC random access disk pool?
I came across one post that said it was supposedly fixed, but recommended
using
a file type diskpool to be safe.


---------------------------------
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!




---------------------------------
Do you Yahoo!?
Yahoo! Small Business - Try our new resources site!


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