Re: Physical layout of a DB in RAC environment

Roman Klesel <rupa_at_firemail.de> wrote in message news:<401A0079.8020808_at_firemail.de>...
> Daniel Morgan wrote:
> > Sybrand Bakker wrote:
> >
> >> On Thu, 29 Jan 2004 08:29:27 -0800, Daniel Morgan
> >> <damorgan_at_x.washington.edu> wrote:
> >>
> >>
> >>> It works just fine. Let the SAN or NAS handle the problem. Once
> >>> you've written to the head ... the rest can just be ignored.
> >>>
> >>> With RAW we got rid of file systems. With 10g even the volume
> >>> management can go away. It is a brave new world ... every 5 years or so.
> >>
> >>
> >>
> >> Do you mean this seriously, or are you just being cynical?
> >> To me the proposed configuration looks like it has been designed by
> >> someone who is either just released from a mental asylum or is a bean
> >> counter (or both)
> >>
> >>
> >> --
> >> Sybrand Bakker, Senior Oracle DBA
> >
> >
> > Seriously.
> >
> > With NetApp or EMC ... the one's I'm familiar with ... once you've
> > written to the head you can just ignore it ... questions about
> > stripping, mirroring, RAID, etc. are irrelevant. The head takes full
> > responsiblity for reading, writing, and recovery.
> >
> > I've also come to the point where I think backing up to tape is a
> > ridiculous waste of time and money. These days you can back up to hard
> > disks far faster and for the same money. Fill a disk, pop it out,
> > replace it with another. Recycle hard disks not tape.
> >
>
> Hmmm...!
>
> So there seems to be a quite spread out spectrum of opinions about this.

:-)

>
> My main 2 doubts in this whole setup are:
>
> 1) The storage System (althogh it has redundant components) remmains a
> single point of failure for all 3 instances: RAC1, RAC2 and shadow.

Find out the vendor's uptime guarantee. If it is orders of magnitude better than any of the other hardware, that's difficult to accept, but once you do, multiply out the failure times of everything else, and worry about the real problems. Which of course, usually involve system monitoring and software bugs.

>
> 2) Does such a storage system prevent I/O contentions effectively
> between REDLOG ARCHIVELOG and DATA/TEMP File aktivity?

It doesn't prevent anything, but certainly moves the bottlenecks about, sometimes in a non-intuitive manner. You can only determine it empirically with some judicious stress-testing.

>
> BTW. We also use RAW devices ... what a hassle ...

That can be fixed easily enough :-)

>
> Regars Roman

jg

--
@home.com is bogus.
http://www.signonsandiego.com/news/uniontrib/sun/business/news_1b1zubkis.html