ADSM-L

Re: SUBFILEBACKUP problems and issues

2003-08-12 06:22:55
Subject: Re: SUBFILEBACKUP problems and issues
From: Steve Schaub <Steve.Schaub AT HAWORTH DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Tue, 12 Aug 2003 06:22:25 -0400
I have been considering using this option on all of our server backups once we 
move to a 100% disk environment for our primary pools.  This would potentially 
save us a lot of disk space, while still allowing for fairly rapid restores.  
But if there are limitations that would prevent us from restoring, obviously 
that will not work.  I agree that the developers should start looking at how to 
enable this option being used at the enterprise level, not just laptop.

Steve Schaub
Systems Engineer
Haworth, Inc
616-393-1457 (desk)
616-412-0544 (numeric page)
4120544 AT Archwireless DOT net (text page)
WWJWMTD


-----Original Message-----
From: Salak Juraj [mailto:j.salak AT ASAMER DOT AT] 
Sent: Tuesday, August 12, 2003 3:26 AM
To: ADSM-L AT VM.MARIST DOT EDU
Subject: AW: SUBFILEBACKUP problems and issues


Hi,

I have a couple of small file servers beeing backed up over VPN´s of 128-256k.

All share same business requirement- to be able to restore single files 
rapidly, while restoration of  whole server - considered to be rare - may take 
a week.

So raising this limits would be of help for me and my network utilisation.

Maybe it could help if some of us  (== more than one) would open 
an official enhancement requirement to IBM?

There had been a small chat with one developer this year on this forum, he 
meant there were no huge technical problems with it.

regards
Juraj Salak



-----Ursprüngliche Nachricht-----
Von: Len Boyle [mailto:SNOLEN AT VM.SAS DOT COM]
Gesendet: Montag, 11. August 2003 23:22
An: ADSM-L AT VM.MARIST DOT EDU
Betreff: SUBFILEBACKUP problems and issues


Hello

We have been using the subfilebackup feature for windows 2k clients located out 
on the wan.This feature has helped quite a bit as the windows client disk usage 
has grown and the wan lines speeds have grown but not at the same rate.

Side note: his is where the tsm client on an icrenmental backup only backups 
changed blocks and not the whole file. This is a huge savings in network and 
tsm server media.To do this the tsm client creates a checksum for each file and 
stores them in a directory on the client.

We have run up against several design limits, which were discussed on this 
listserv a few months back. Now we have run into a problem that appears to be 
the same as listed in apar IC36552.

Both the limits and the problem are related to that magic 1 and 2 gig numbers.

The problem is when the base file and delta files, each under the 2 gig support 
limit, in total exceed 2gig. Then the client can not put the data together to 
restore the file.It appears that the client uses a 2gig memory map of the data.

The first limit is that the files used to store the checksums have a limit of 1 
gigbyte.

The second limit is that the files covered under subfile are limited to 2 gig.

The orginal design thoughts for this feature seemed to be for laptops with 
small disk drives. Of course we are using this for windows files servers with 
users who store files greater then 2gig.

So we run run up against both of the above limits. And now we are finding that 
the file limit is really going to have to be something much smaller then 2gig, 
because of apar ic36552.

Can I ask if there is a test fix for Apar IC36552, that I be put on the list of 
folks that are on the testors list.

Also let me put in a vote to raising the limit on the file size and the 
checksum size.

How many folks out there are using this feature?
For servers?
On the lan and not the wan?


-------------------------------------------------------------------------
Leonard Boyle                               snolen AT vm.sas DOT com
SAS Institute Inc.                          ussas4hs@ibmmail
Room RB448                                  len.boyle AT sas DOT com
1 SAS Campus Drive                          (919) 531-6241
Cary NC 27513

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