RE: RMAN Performance Maladies

  • From: "Michael Fontana" <MFontana@xxxxxxxxx>
  • To: <naqimirza@xxxxxxxxx>, "Jared Still" <jkstill@xxxxxxxxx>
  • Date: Tue, 13 Feb 2007 14:48:15 -0500

OK - on the raid 5 issue - just to be sure everyone knows I am comparing
apples to apples:
 
SQLBACKTRACK was writing to the same raid5 san mount point.
 
Jared hit on something that will cause me to do some more testing - and
that is that whilst sqlbacktrack merely backs up the datafiles to disk
(a file-based backup), rman is checking blocks, etc.
 
Someone also suggested setting up RMAN to do a file-based backup.
Haven't seen that documented anywhere, but if it's an option, we'd
certainly consider it.
 
On the issue of SLA - I agree it was set a long, long time ago when our
largest databases were about 60m, and probably should be revisited, but
our boss has not yet given us permission to go to incrementals until we
know all of our options here.
 
One thing that is not an apples to apples comparison that we need to
test: We've never backed up THIS database with SqlBackTrack after
migrating it to 10g 64-bit.  I will let the list know how that turns
out.
 
 

________________________________

From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Naqi Mirza
Sent: Monday, February 12, 2007 11:59 PM
To: Jared Still
Cc: oracle-l@xxxxxxxxxxxxx
Subject: Re: RMAN Performance Maladies



This is true (RAID5:READS), the strange thing , though, is there is also
a reporting server - which is not an entire clone of this database - but
pretty close to it.
If users are asked to disconnect from the reporting server (basically no
load) and a test backup is performed via rman - the throughput achieved
is 50mb/s. The storage (that houses the database) is configured in the
same manner, also uses RAID5. The backup is to the same tape drives. I
understand the significance of a number like 666, but whats up with 50.
Is it possible, though, when a large number of disks (100+) are part of
a RAID5 configuration for the controllers involved to become bound with
i/o requests?
Sorry , for the slight hijack of your thread Michael.


----- Original Message ----
From: Jared Still <jkstill@xxxxxxxxx>
To: naqimirza@xxxxxxxxx
Cc: MFontana@xxxxxxxxx; oracle-l@xxxxxxxxxxxxx
Sent: Tuesday, 13 February, 2007 5:02:08 AM
Subject: Re: RMAN Performance Maladies




On 2/12/07, Naqi Mirza <naqimirza@xxxxxxxxx> wrote: 


        The problem I witnessed at a certain site - never got resolved,
the site was also using RAID 5 - I suspected this as the culprit 
        


I would not expect RAID5 to be a problem during backup, unless there is
a 
lot of write activity to the drives during that time. 

-- 
Jared Still
Certifiable Oracle DBA and Part Time Perl Evangelist



________________________________

Web email has come of age. Don't settle for less than the All New Yahoo!
Mail
<http://us.rd.yahoo.com/mail/uk/taglines/gmail_com/nowyoucan/web_mail/*h
ttp://us.rd.yahoo.com/evt=40566/*http://uk.docs.yahoo.com/nowyoucan.html
> .

Other related posts: