« dNFS Performance Reprise | Main | "Blended" FCP / NFS Oracle Solution (Reprise) »

February 18, 2008

Comments

Alex McDonald

Jeff, I know a lemon when I see one. Honestly, EMC can't be serious with this, can they?

"I have not done an exhaustive search, but as near as I can tell, no other storage vendor has done this yet."

Little wonder. Why would anyone else do this, for goodness' sakes. It's terrible! There are so many cogs and wheels in this faux "solution", not to mention the expense, footprint, power consumption, FC and IP management, and the different software for the SAN and the NAS. Not very practical, is it?

(BTW, the Celerra appears to be oddly cabled to the CX. Perhaps I'm being picky here, but it doesn't look right?)

"the LUN that you see over FCP is a really, live, good, old-fashioned LUN"

And the advantage of that is? Presumably you don't really take to VMware either, because you can't see the "really live, good, old-fashioned server" there either.

I'm stumped. What exactly is this demonstrating? There's no amount of blending going to fix the fact that the lumps in this idea are just too big to whizz away.


Disclaimer; I work for NetApp, where we have Oracle stuff that just works. But you know that.

---------------------

Response by TOSG:

Alex, I don't know you so perhaps you started working at NetApp after I left. However, I am not sure what "cogs and wheels" you are referring to.

The Celerra is shown as being cabled to a CLARiiON because that's exactly how a Celerra works: It is a front end to a CLARiiON. You have heard of the gateway filer, right? That's the same concept as the Celerra. A head sitting in front of a world-class SAN storage array.

Further, in terms of the amount of software that must be installed, it actually less than a pure FCP solution. In the case of FCP, you need to configure a shared storage layer for the CRS layer, i.e. the voting disk and OCR files. That requires either OCFS2 or raw.

With NFS, it is simply built-in.

A Celerra already includes a CLARiiON backend, and we are simply using the available ports on the piece of equipment that is already sitting there, i.e. no added equipment. We are installing less software on the database server.

So, again, what cogs and wheels are you talking about?

I can tell you that I have installed and configured Oracle RAC dozens of times, on a large number of environments. Doing it this way, i.e. FCP for datafiles, tempfiles, logfiles and controlfiles, and NFS for everything else, i.e. archived logs, CRS files, backups and flashback is far, far more manageable than a pure FCP solution, while providing identical performance. That's why a customer would want to do this.

Have you actually installed and configured Oracle RAC in an FCP enviroment? Really, I'm interested.

The comments to this entry are closed.

Powered by TypePad
View Jeff Browning's profile on LinkedIn

disclaimer: The opinions expressed here are my personal opinions. I am a blogger who works at EMC, not an EMC blogger. This is my blog, and not EMC's. Content published here is not read or approved in advance by EMC and does not necessarily reflect the views and opinions of EMC.