Re: Database layout - file distribution

Daniel Morgan <damorgan_at_x.washington.edu> wrote in message news:<1071848441.531489_at_yasure>...
> Milan Smederevac wrote:
>
> > Hi,
> > I have a question:
> > I'm trying to figure best possible physical layout but as always we
> > are one disk short for optimal layout, I've read K.Loney DBA handbook
> > on that subject
> > and there is suggested combination of RBS and TEMP segments but would
> > it be mistake to put TEMP segments with redo log files?
> >
> > So, what is better, or less bad. To combine RBS wit TEMP tablespace on
> > same HDD
> > or to have TEMP datafiles combined with redolog files ?
> >
> > I think that is pretty the same, RBS and Redo logs are write intensive
> > when we have some updates or deletes, on the other side TEMP segments
> > are used when we have some sorts that can't be done in memory?
> >
> > thanks for reply in advance
> > M.S.
> >
> > PS 8i on SunOS 5.6
> > with 10 mirrored HDDs dedicated for DB.
>
> The advice you are reading is old and no longer applicable to most
> systems. Talk to your SA and ask if they can stripe and mirror
> everything (SAME).

Juan Loiza ppt presentation on this:
http://otn.oracle.com/deploy/availability/pdf/OOW2000_same_ppt.pdf http://otn.oracle.com/deploy/availability/pdf/oow2000_same.pdf

Personally, I'd always isolate the log and archived redo log files. Use at least 2 members per group, on separate RAID volumes, on separate controllers if possible.

An 8 drive RAID 10 volume for datafiles is a good start. You'll want the stripe size to align with the operating system's max IO size.
Check Connor McDonald's site for info as to how to determine this (I can't find his site anymore - where did it go?).

If you have a mixed environment of both DSS and OLTP, you might want to consider having more than one mount point for datafiles. I prefer to have the segements for the oltp part of the app not stored with the segments that are hit via parallel query option, so as to help ensure (heuristically) that the response time degradation on the oltp transactions is minimized.

there is an article on otn regarding Automated Storage Management:

http://otn.oracle.com/oramag/webcolumns/2003/techarticles/scalzo_asm.html

I saw a presentation on this at NYOUG about three months ago that touched on this, but it did not cover it in detail.

http://www.nyoug.org/presentations.htm
http://www.nyoug.org/sarwal_10g.pdf

a query against a search engine (e.g. google) that I used was:

oracle automated storage management asm 10g

found this:

http://otn.oracle.com/oramag/webcolumns/2003/techarticles/murphy_asm.html

From what I picked up, you'll have filesystems for OS, binaries, config, log, trace files, and you'll turn over disks to Oracle to handle. Whether these are physical drives, RAID volumes as presented by the hardware RAID contollers or mount points on a SAN should not really matter.

Software mirroring (in oracle) of the system tablespace's datafiles might seem attractive (on top of hardware mirroring) but I wonder what the additional overhead of having multiple datafiles to read from would have on the consistent read mechanism, I would guess none.

Not being able to access files from the host operating system is a little troubling. One would have to have extreme confidence in RMAN prior to making that leap.

Pd Received on Sat Dec 20 2003 - 12:12:23 CST