Re: RAID strip size and db_file_multiblock_read_count

My point is that wouldn't it be faster to have several disks in the array servicing each logical read in parallel, than to have each physical read serviced by only one disk. The idea is that 4 16k reads done in parallel is faster than 1 64k read done serially.

I think I may have been confusing the terminology though. As it turns out, I wanted to say "segment size" instead of "stripe size". Segment size is the size of the stripe on an individual disk. Stripe size is the total stripe size (num disks * segment size).

In article <36FFFACF.24B9DDAB_at_rdbms.freeserve.co.uk>,   "Christopher M. Day" <christopher.day_at_rdbms.freeserve.co.uk> wrote:
> Chuck,
>
> 1. Yes, you want to maximise the sequential reads, rather than incurr
> the performance penalty by moving the heads.
>
> 2. Sorry I don't follow your logic. You imply that it is better to
> perform more physical reads for one logical read, than have one physical
> read for one logical read.
>
> As regards table scans, I would try to get the disk subsystem to do as
> many sequential reads as possible (fastest), couple that with
> tablespace and partition striping and the PQO.
>
> Chris.
>
> chuckh_at_safeplace.net wrote:
> >
> > The ORACLE tuning guide says that the minimum strip size to use in a raid
> > array is twice that of db_block_size * db_file_multiblock_read count? Why is
> > this? Won't that force each read for a table scan to be limited to one disk
> > in the array? I'd think that you'd get better performance by making it just
> > the opposite and so that each table scan read is done accross multiple
disks.
> > -- Chuck Hamilton chuckh_at_safeplace.net
> >
> > If at first you don't succeed, skydiving isn't for you!
> >
> > -----------== Posted via Deja News, The Discussion Network ==----------
> > http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
>
>

--
Chuck Hamilton
chuckh_at_safeplace.net

-----------== Posted via Deja News, The Discussion Network ==---------- http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own Received on Thu Apr 01 1999 - 09:03:13 CST