[Veritas-bu] Restore performance questions in NB

[spaldam]

I've dealt with this exact issue many times before.  It’s a balancing act that 
really boils down to this:

Only one restore job can access the same tape at a time, and each restore job 
can only use one tape at a time (there is a new feature for doing parallel 
restores that were multiplexed, but that only applies to certain situations 
that databases don’t fall under).  The best solution for this is to duplicate 
all the images on the tapes to a DSU (disk storage unit), then all the restore 
jobs can access the DSU all at the same time. Make sure you duplicate the 
entire tape (not individual images) to get the duplication done as quickly as 
possible.  This is also one of the reasons why having your on-site backup copy 
be on disk is getting to be so popular.  In fact, I wouldn’t even consider 
doing hourly database log backups directly to tape.

Also, the less often you do your log backups, and the more logs you get backed 
up in a single stream (meaning fewer backup images) the faster your restores 
will be.

To answer your questions:

1-2. Volume Pools can help you divide up your data and improve performance, if 
done correctly, but it can also cause you to use more media with tapes not 
getting filled up.  Again, it’s a balancing act, but most place I’ve worked at 
are more concerned about costs than performance (at least until they need 
something restored ASAP).

3 & 5. You can only specify storage units for backups (not specific tape 
drives), so you have to configure your storage units accordingly if you want to 
restrict which drives get used for what.  On the other hand, restores only care 
which server they were backed up on (I'm not familiar with a way to force the 
use of a certain storage unit for a restore).  You may want to make a close 
inventory of your storage units to see why you always have a couple of drives 
available; it may have been designed that way on purpose to allow for restores 
or user initiated backups to be done at any time.

4. Especially with Databases, it's a delicate balancing act.  Improving one 
often has a negative impact on the other; again using disk staging areas or VTL 
can help a great deal with this by eliminating the need for multiplexing and 
improving overall performance.  Just remember tapes are linear, so to get to 
the data at the end of the tape, means the drive has to pull the entire tape 
through it, and it can only read one part of the tape at a time.

+--
|This was sent by [EMAIL PROTECTED] via Backup Central.
|Forward SPAM to [EMAIL PROTECTED]
+--


___
Veritas-bu maillist  -  Veritas-bu@mailman.eng.auburn.edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu

[Veritas-bu] Restore performance questions in NB

[rascal]

Good day all,

So here is an interesting question regarding restore performance.  First,
I'll lay down the scenario and then I'll ask the question to help give all
of you Netbackup Gods a good base. {:-P

Right now, I have a configuration like this:

Master = solaris 10/gb nic config w/ matching media servers
fibre connected to stk9310 w/9940Bs'
clients = mostly aix w/ oracle 9i-10gr2

for every database, there are two policies created:

daily full backup
hourly logs

the full backup goes to a set of tape drives and a set of media.  The tape
drives are broke into storage units based on the media server and the media
is a large pool.

The logs go to disk and then are duped off to tape at a later time.

Here are my problems:

1.  Drive contention:  During a sample DR exercise, I launch multiple DB
restores and run into restores waiting on 1)drives 2) media
2.  Restores run slow:  I have looked at network and that has been tuned out
pretty well but I think this boils back down to problem 1

Here are my questions:

1.  Since my databases all touch the same tape drive pools, is there anyway
to break that down a little further to gain performance on restores?
2.  If I broke out the media pool more, say on the grades of the DB's being
backed up, would that help performance?
3.  Is there a way to tell Netbackup which set of drives to use during a
restore and if so, can it be setup somewhere in the policy or just during
the restore?
4.  Is there a better way to configure policies such that I can get good
backup/restore performance without sacrificing both?
5.  Drive contention is always an issue; I have multiple drives available
and it seems like even at peak performance, backup or restore, I have a
couple of drives sitting idle; any ideas on how to fix this?

Again, thanks for any suggestions/pointers/things to read/prayers/etc...
that can be given on this.

-- 
Matthew MCP, MCSA, MCTS, OCA
[EMAIL PROTECTED]

Define Trouble:
Why did you apply THAT patch??
___
Veritas-bu maillist  -  Veritas-bu@mailman.eng.auburn.edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu