RE: RAC configuration Sun Cluster v. Veritas

  • From: "Hameed, Amir" <Amir.Hameed@xxxxxxxxx>
  • To: <smcclure@xxxxxxxxxxxx>, <oracle-l@xxxxxxxxxxxxx>
  • Date: Wed, 1 Jun 2005 13:25:16 -0400

Hi Steve,
I am in the process of implementing a three-node RAC using Oracle 9i2
(9.2.0.6) and 11i apps on Solaris 8 in my lab environment. In the lab, I
am using three v880s, but in production, we will be using two 20-CPU
domains from E15K and one domain from E12K (a total of three domains).
When we decided to go to RAC about 1.5 years ago, I did a lot of
research and compared benefits of Sun cluster versus Veritas DBE/AC for
9i RAC. DBE/AC was a newer product as compared to Sun cluster at that
time and did not really have a very large client based as compared to
Sun (for OPS). The main thing that had attracted me to Sun was their Sun
Firelink interconnect that was based on the RSM (Remote Shared Memory)
technology. The reason it attracted me was because it's latency was very
low (in the range of micro-seconds) and no other interconnect in the
market came even closer to it. But then I learned that there were not
very many implementations of this interconnect out there in the field.
Sun did not have any clustered filesystem at that time and we had to
choose RAW if we decided to go with Sun. Sun rep. initially tried to
sell us the Sun cluster by stating that their GFS (global filesystem)
was certified by Oracle for RAC but that was not the case. We were
already using EMC Symmetrix 8830 SAN at that time and already had the
backup and all procedures in place. Using RAW would have meant altering
some of those procedures. Since we were using 11i applications, the
database was comprised of over 550 data files at that time (now they are
close to 650 files and 750 GB in size) and that would have meant
converting all those filesystem-based files to RAW files.
We have also been very pleased with Veritas's support so far. If you are
going with DBE/AC or their foundation suite, you may want to be aware of
a bug that is related to ODM. The bug causes the MBRC IO to be split at
the ODM layer into multiple single blocks. So, if your MBRC is set to 8
and you have a 8k block size, ODM will break down the IO into eight 8k
blocks. There is a patch available for 10G, 9.2.0.6 and 9.2.0.4 (Oracle
had back ported the patch for us from 9.2.0.6 to 9.2.0.4)

Thanks
Amir

-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Steve McClure
Sent: Wednesday, June 01, 2005 1:17 PM
To: oracle-l@xxxxxxxxxxxxx
Subject: RE: RAC configuration Sun Cluster v. Veritas

>>Steve,

>>You've narrowed it down quite nicely in your requirements.  You
mentioned
>>"either Oracle9i or 10g", and that is the key ambiguity, so to speak.


I should have clarified that a bit more.  My thinking was that we would
use
9i unless we went with the Sun Cluster and had to go with (or so I
thought)
10g for file management options.  I just didn't see a way to break
things
down as clearly as you did when I was posting.

It seems that the Veritas Storage Foundation solution makes a lot of
sense
in that it is a more mature product than the Sun/Oracle solutions.  I'm
starting to consider the Veritas configuration as the way to go, unless
our
research into the other approaches shows one of them to be clearly
superior
for our exact situation.  A former manager said that it is great to be a
pioneer, but you have to keep in mind that a lot of pioneers ended up
with
arrows in their backs. Proven solutions do have their merits.

Thanks for the help,

Steve McClure

--
//www.freelists.org/webpage/oracle-l

--
//www.freelists.org/webpage/oracle-l

Other related posts: