Re: 10g ASM is junk ...

Tony Dare <tonydare_at_yahoo.com> wrote in message news:<IM2tc.22$Bk6.58_at_news.oracle.com>...
> Please excuse me, I am new to the news group and don't mean to intrude,
> but as I understand it ASM stripes your data across Logical Units, not
> disks per se. So you implement hardware RAID as you like create your
> volumes (LUNS) and then Oracle ASM manages data placement across the
> logical units and rebalances according to amount of data and intensity
> of usage. This is what I have read. Those who have taken offered
> classes on it would have the straight "poop," of course.
>
> My $0.02 only.
>
> Cheers and thanks,
>
> Tony Dare

Hi Tony.  

From what I read in the Oracle Press New Features Guide for 10g, the DBA/SA can choose, when managing database storage (CREATE DISKGROUP) as to whether to have oracle perform mirroring across "disks" (default is for a mirrored set) or to tell ASM that the external storage is already mirrored externally and that you do not want to have FAILGROUPS for that DISKGROUP.  

wire diagram for CREATE DISKGROUP
http://download-west.oracle.com/docs/cd/B13789_01/server.101/b10759/statements_5008.htm#SQLRF01114

CREATE DISKGROUP diskgroup_name
  [ { HIGH | NORMAL | EXTERNAL } REDUNDANCY ]   [ FAILGROUP failgroup_name ]
DISK qualified_disk_clause

       [, qualified_disk_clause ]...
    [ [ FAILGROUP failgroup_name ]
DISK qualified_disk_clause

           [, qualified_disk_clause ]...     ]... ;

If you have the (physical) hard drives and controller channels available, hardware mirroring and ASM FAILGROUPS would be desirable, depending upon how valuable the data and availability are to the business.

I haven't yet kicked the tires on ASM personally, nor do I know what "fine" vs. "coarse" correspond to in terms of physical stripe size.

There is a table of recommended disk group configurations here: http://download-west.oracle.com/docs/cd/B13789_01/server.101/b10739/storeman.htm#i1007979

the entire 10g documentation set can be downloaded from here: http://otn.oracle.com/documentation/database10g.html

From the DBA standpoint, the idea that no one else other than ASM can touch these physical hard drives is of great benefit, as you're not going to be hunting who is performing non-oracle IO against them. But the idea that backups are entirely managed by RMAN doesn't leave me feeling warm and fuzzy.

hth, and apologies for not having read the 10g concepts guide yet.

Pd

>
> Mark D Powell wrote:
> > "Niall Litchfield" <niall.litchfield_at_dial.pipex.com> wrote in message news:<40b3973a$0$20508$cc9e4d1f_at_news-text.dial.pipex.com>...
> >
> >>"Domenic" <domenicg_at_hotmail.com> wrote in message
> >>news:c7e08a19.0405241711.3c55f744_at_posting.google.com...
> >>
> >>>I've been looking over ASM in 10g. Am I the only one who thinks ASM
> >>>is a piece of junk? It looks like software-based RAID, and you need
> >>>to take on the overhead of an ASM instance. How can this possibly be
> >>>faster or better than striping/mirroring at the operating system
> >>>level?
> >>
> >>I don't see it as junk, and I don't see it as software RAID - I see it as a
> >>software SAN (or possibly a cheap SAN). I'd rather have a hardware SAN and
> >>none of this auto balancing malarkey (I believe that is the technical term
> >>:( ) .
> >>
> >>lets see what it looks like in 10g release 3 or 11x release 2.
> >
> >
> > Apple Computer has just come out with a software SAN. I think it is
> > called xSAN. There may be some advantages of software over hardware
> > in this case. Quite a few sites have had problems properly
> > configuring their hardware SANs. A software version can take user
> > errors out of the equation.
> >
> > I think it may be too early to judge the merits of ASM. The question
> > right now may be just how good a job has Oracle done on the
> > logic/code? For sites with no real Oracle DBA or DBA's forced to work
> > with several databases at once without the benefit of in-depth
> > knowledge it may be the best disk solution.
> >
> > IMHO -- Mark D Powell --
Received on Thu May 27 2004 - 14:10:07 CDT