Message: 4
> From: s.pabbatireddy AT ieee DOT org
> To: veritas-bu AT mailman.eng.auburn DOT edu
> Cc: s.pabbatireddy AT ieee DOT org
> Date: Wed, 7 Mar 2001 08:47:20 -0500
> Subject: [Veritas-bu] Veritas Oracle backup solution: BLIB 
> versus RMAN - Please help!!!
> 
> All:
> 
> I posted a message earlier on this topic, but unfortunately haven't
> received any responses from the group. Trying again!!
> 
> We have purchased netbackup 3.4 as our enterprise backup solution. Now we
> are planning to rollout the solution for oracle databases. Now we are
> debating between BLIB versus RMAN, and need your help in comparing both of
> this approaches.
> 
> Our environment:
> - Oracle database versions 7.3.4/8.0.5 on hp-ux 10.20

RMAN in 8.0.5 is seriously broken.  The fix is to patch Oracle up to 8.0.6
or higher.  The issue is documented in Metalink somewhere.  I found out when
by backups started hanging and I called Oracle support.  

I was running on Solaris so I suppose it is possible the problem is related
Oracle-on-Solaris.  You can check.

> - Oracle database version 8.1.6/8.1.7 on hp-ux 11.0
> - OLTP and Dataware house environment
> - master server NT
> - Media server NT
> - Netbackup 3.4
> 
> Our Goal (in the order of priority):
> - Our online application performance should not be impacted by the backup
solution
> - Optimal backup/recovery performance
> 
> I am looking for answers to the following questions:
> 1. Does BLIB degrade online application performance to maintain the
> modified blocks through storage check point. If so, how much?
> 2. We refresh around 30G data to our datawarehouse environment by-weekly.
> Is the BLIB a right solution for the databases with massive data changes?
> 3. How much additional disk space is needed to go with the BLIB solution.
> Our environment is very dynamic, some times we see 50MB log switches every
2 minutes.
> 4. How can we compare the backup/recovery performance between BLIB and
> RMAN/EBU. ( I have seen, RMAN is optimal for fullbackups). If you have any
> bench mark numbers, it will help us a great deal.

I'm not sure how to respond to this assertion.  I'm not an expert DBA nor do
I have extensive experience with RMAN but I have seen Legato (not Netbackup)
hot backups go from 3 hours under 7.3 where the technique was table spaces
in hot backup mode to 20 minutes under 8.0 and RMAN hot backups.  Both
backups were daily fulls.  Oracle claims that RMAN reads the database blocks
and skips empty ones, even on a full backup.  YMMV, 'natch.

> 5. Is BLIB supported for HP-UX 10.20 environment with oracle 7.3.4, 8.0.5
with netbackp 3.4?
> 6. If RMAN is used, where does all the information required for recovery
is
> stored (in the RMAN catalog or master server catalog)?

We are using RMAN 8.1.6 without a recovery database.  My DBA tells me this
works just fine but our options on recovery are limited putting everything
back exactly as it was during the backup: same locations and file sizes.
Since our database is sized very large but still 90% empty, this could
impact the speed of recovery but we've committed to restoring to similar
hardware so we're OK with it.  If you use the recovery database, RMAN can be
scripted to do some pretty amazing things since it keeps a lot of
information about the database and the backup.  Netbackup is treated as if
it were a file system: files are named and written to it or requested and
read from it.  RMAN handles all the other gritty details.  Without a
recovery database, information about the backed up database is kept in the
control files.  Fewer options are available during recovery - things like
resizing the database, for example, or renaming data files.
 
The downside to a recovery database is that you have to license it and you
can't use RMAN to back it up.  For one production database, Oracle suggests
putting it on the production system with it's own disk hardware to minimize
shared failures.  For two production databases, use two recovery databases
on each other's hardware.  Beyond that centralize.  If you keep an open
window for it, you can close the recovery database and do a cold backup.
That's fast and easy and scriptable.

> Your input and help is highly appreciated.
> 
> Thanks in advance.
> 
> Sudheer
> ====================================================
> Sudheer Pabbatireddy
> Manager, Database Systems
> IEEE, Inc
> Phone: (732) 562-6534
> Fax: (732) 562-1727
> Email: s.pabbatireddy AT ieee DOT org
> ====================================================

Dana Bourgeois
Slam Dunk Networks 
Digital Mechanic & Network Janitor
1.650.632-5543
1.650.996-5687  [cell]