Re: db_file_multiblock_read_count value

• maxphys is 1 mb on solaris ( or has this changed ??? )
• we set the stripe with to 2048 with veritas volume manager, with equals to 1M, to improve the performance, (minimal latency ) this is very important

think how ridiculous it is to have a transfer of 8MB and a stripe of 64kb because the head movements are 10000 times slower than the time you can save with a big transfer size , so the stripe width and the transfer size need to be the same and big transfer only helps with big stripe width ( no head movement during transfer )

• i would go to 16kb blocksize , so db_file_multiblock_read_count is 64

do you use veritas 3.5 with ODM ??

we use DBE 3.5 with ODM ( replacement for quick io )

only 3.5 is certified with 9.2 as I know.

I try to find a shop who also works with this setup, so perhaps you do

"danilo" <daniloa_at_bbirger.intnet.mu> schrieb im Newsbeitrag news:9713eeb.0210310426.17a11a7_at_posting.google.com...
> Hi,
> I would like to know why we can't set Oracle parameter
> db_file_multiblock_read_count greater than 128 on a Sun system running
> Solaris 8 (with latest patches) and Oracle 9i release 2.
>
> On the OS side, we set the max. physical I/O size(maxphys) to 8 M in
> /etc/system and sd_max_xfer_size is also set to 8M in the
> /kernel/drv/sd.conf. We tried to set db_file_multiblock_read_count to
> 1024 as recommended by VERITAS (maxphys/Oracle_block_size) = 8 M/8 K
> but Oracle just seems to retain the 128 value.
>
> Does anyone know if this is an Oracle limitation or something else?
>
> Thanks
>
> Danilo
Received on Thu Oct 31 2002 - 09:03:43 CST