ADSM-L

Re: [ADSM-L] Client side dedup and API clients

2011-02-07 23:06:31
Subject: Re: [ADSM-L] Client side dedup and API clients
From: Dave Canan <ddcanan AT GMAIL DOT COM>
To: ADSM-L AT VM.MARIST DOT EDU
Date: Mon, 7 Feb 2011 20:05:44 -0800
Steve, there is a way to do a simple trace on the client side to see if
client side dedup is working. I don't have access to the system but I
will send an email with details tomorrow . I referred to this technique
in an IBM presentation I did last year on TSM V6.2 client deduplication.

Dave Canan
IBM Advanced Technical Support
DdcananATUSDOTIBMDOTCOM From: Steve Harris
Sent: Monday, February 07, 2011 7:19 PM
To: ADSM-L AT vm.marist DOT edu
Subject: [ADSM-L] Client side dedup and API clients
 Hi All

 How does one determine whether client-side deduplication is working for
 API clients?

 I have a couple of test systems, a TSM 6.2.0.0 server (UNI01) and a
 6.2.2.0 server (UNI00).  UNI01 has been backing up its database, server
 to server to UNI00 for 6 months or so.  After upgrading UNI00 to 6.2.2 I
 turned on deduplication for the server-to-server pool used for DB
 backups.  The identify process did its thing and I saw deduplication
 happening in the pool.

 After a week or so of this I have created a client option set on UNI00
 for UNI01 and turned on client side deduplication.
 The source (UNI01) db backup summary record shows the same sort of
 volumes as it did before the change.
 There is no summary record on the destination (UNI00) that correlates
 with the source DB backup.  There are accounting records but these show
 no difference, and the archived data and network data are the same size
 within a few bytes.
 There is no actlog message on the source side to show that the DB
 backup was deduplicated or what the deduplication benefit might be.
 despite that, the amount of work that the identify process is doing per
 backup seems to be less than it was before client side deduplication was
 turned on, which is what would be expected.

 I need this to work before 6.2 goes full production as I have two
 datacentres some distance apart with a small but adequate pipe between
 them. 5.5 full/incremental server to server DB backups work fine, but
 incrementals aren't available in 6.2.

 So the specific question is how do I tell if client-side dedup is
 working in the server-to-server case.  But, more generally we have three
 sources of information about client backups: Actlog, summary and
 accounting.  BA clients dump their dedup stats to the actlog, but API
 clients don't have that luxury. Summary and Accounting record formats
 don't seem to have had this information included.

 Regards

 Steve.

 Steven Harris
 TSM Admin, Paraparaumu, New Zealand

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