Re: ASM setup
Date: Mon, 28 Apr 2008 13:35:44 -0700 (PDT)
On Apr 28, 2:57 pm, "Vladimir M. Zakharychev" <vladimir.zakharyc..._at_gmail.com> wrote:
> On Apr 28, 9:39 pm, macdba321 <macdba..._at_gmail.com> wrote:
> > On Apr 28, 12:45 pm, hpuxrac <johnbhur..._at_sbcglobal.net> wrote:
> > > On Apr 28, 10:26 am, macdba321 <macdba..._at_gmail.com> wrote:
> > > > Hi all,
> > > > I'm setting up my first ASM based database ever. (Just for testing
> > > > and learning, NOT for production). I only have 3 physical disks in the
> > > > ASM disk group. (4 total physical disks in the server).
> > > > I am loading the OS (Windows 2003 server) on one physical disk as
> > > > well as the Oracle installation. I am keeping the other 3 disks raw
> > > > for ASM.
> > > > Should I RAID the 3 raw disks (hardware RAID and/or Oracle raid) or
> > > > just let Oracle do what it wants with them?
> > > > Thanks!
> > > What most people do in the real world ( the ones actually using ASM
> > > anyhow ... not a large percent yet of the oracle installed base ) is
> > > basically this.
> > > If your storage environment supports RAID well then they get LUNs from
> > > it and define the ( LUNs become disks in an ASM diskgroup ) diskgroup
> > > as external redundancy.
> > > This tells ASM hey don't bother doing any overhead of mirroring the
> > > extents you allocate across disks ... my storage environment does a
> > > nice job of dealing with that stuff.
> > > If the storage environment doesn't support RAID then they point ASM at
> > > disks and let ASM do the mirroring, You have a choice of 2 or 3 way
> > > mirroring to add additional protection levels.
> > > So all in all this design is up to you ... perhaps it should be based
> > > on the type of testing that you want to do and how that relates to an
> > > eventual production environment that you might be running in
> > > eventually.
> > > There's a book that you might want to think about getting "Oracle
> > > Automatic Storage Management" it's one of those Oracle Press books
> > > authors are Vengurlekar, Vallath and Long includes 10g and 11g.
> > > Personally I recommend getting and reading that book ( it really isn't
> > > that long ) and spending a day or 2 reading it before you make more
> > > decisions about your setup.
> > Hello, apologies if you receive this reply twice (It didn't reply
> > successfully the 1st time).
> > I just got the book you recommended and am playing with the database
> > server as I go along for fun (since for once I have time on my hands).
> > The database server has 3 raw disks for me to "play" with.
> > This is 100% academic/fun. There is ZERO need for High Availability/
> > redundancy. If I lose a drive, oh well, I start over. It doesn't
> > matter.
> > One quick question:
> > If you wanted to set it up for 100% performance/speed, would you:
> > A) Hardware-based RAID-0 the 3 raw disks as 1 ASM disk group,
> > B) Oracle-based RAID-0 the 3 raw disks for ASM
> > C) Some other RAID level?
> > D) Other???
> > Thanks!!!!!!!
> I'd do D): one spindle for ASM DG1 dedicated for redo logs and
> possibly a copy of control file (yes, it's going to be tremendous
> waste of space, but you wanted speed, didn't you? ;) and two spindles
> striped with hardware RAID-0 for ASM DG2, where the database (in
> NOARCHIVELOG mode!) and flash recovery area would be. Both disk groups
> would be with external redundancy to minimize mirroring overheads;
> separate spindle dedicated for redo should minimize the i/o component
> of redo log sync times. This setup automatically offers zero data
> protection as requested. :)
> My $0.02,
> Vladimir M. Zakharychev
> N-Networks, makers of Dynamic PSP(tm)
Thank you. I was starting to have the exact same though as you. (1 ASM group comprising 1 disk for redologs & control + 1 ASM group comprising 2 disks for datafiles and flash.) Seems like the fastest possible setup.
What I don't know however is if the 4 disk share 1 controller, 2 controllers, etc??
I guess regardless, that must be the fastest setup!
Thanks again. Any other thoughts?? Received on Mon Apr 28 2008 - 15:35:44 CDT