How do you lay out your Oracle filesystem structure?

[Michelle Barton]

We're reviewing the way we have laid out our oracle filesystem structure now that we are getting an XP24000 in order to determine what we need now :) We currently have 3 /u*/oralog directories, all in different vg's. We have a seperate /u*/oraarch directory in it's on vg. And then we have /u*/oradata.

We'd like to combine this to some level. Thoughts, ideas, best practices????

Thanks!

[TwoProc]

Actually Michelle,

If I gave you one of my standard documents - I'd tell you to separate them even further.

I'd have redo logs on their own area,
I'd have undo logs on their own area,
I'd have system data files in it's own area,
I'd have report output on it's own area,
I'd have the remaining data files spread across several areas as well, depending on what purposes (read as loads) you've got going on with the server. I can assume since you're putting it on an XP24000, it's large.

So, I'd say you're going the wrong way.

[TwoProc]

You can find my previous Comments on Disk layout for Oracle. While I realize you're wanting to go the other way, there is discussion in there on which way to go to reduce mount points from the full blown version; so you may find some of it relevant.

http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=981200

Another:
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=1170408

I would say that you only need to /u0*/oralog directories, not three, as long as you interleave them. Make sure that they alternate order. For example,

on mount point A have redos 1 3 5 7 9
and on mount point B have redos 2 4 6 8 10

That way, at most, at any point in time, you're writing to one redo log as the active redo, reading from the other redo log area to write its data to the archive log. Of course, this would happen as you're switching from say, redo 3 to 4 and starting your archive from redo 3 to the archive log disk area. I don't see that you'd need three areas to get this done. And, if you're not interleaving them, the you're not accomplishing anything more than a single mount point would do for you anyways, except at cross over points, wherein the last redolog on the first mount point is being switched out for the first one in the second mount point, for example.

Hope this helps somewhat anyways.

[Yogeeraj_1]

hi,

Allow me to also add.

ASM for your datafiles would equally be worth considering.

good luck!
kind regards
yogeeraj