topic Re: db_block_size in Operating System - HP-UX

db_block_size

YLTan — Fri, 01 Nov 2002 03:54:10 GMT

what's the UNIX kernel parameters equivalent of db_block_size in Oracle DB ??

Re: db_block_size

Yogeeraj_1 — Fri, 01 Nov 2002 05:53:35 GMT

hi,

from metalink:

To display the file system block size, execute the "tunefs" command as root:

# tunefs -v

is the name of the device you would like to display the file system block size of.

For example:
# tunefs -v /dev/dsk/c2t6d0

*NOTE: tunefs is not necessarily installed on all HP/UX systems.
tunefs will only work for HFS, and not for JFS.

Hope this helps!

Regards
Yogeeraj

Re: db_block_size

T G Manikandan — Fri, 01 Nov 2002 06:03:06 GMT

check for the man pages of mkfs_hfs and mkfs_vxfs.

check for the blksize option

YOu have to specify the block size during the creation of file system.

Like

hfs file system
# newfs -F hfs -b xxxx /dev/vgxx/rlvyy

# newfs -F vxfs -b xxxx /dev/vgxx/rlvyy

You can check the existing block size using the

#fstyp -v /dev/vgxx/rlvyy

Thanks

Re: db_block_size

Gary Yu — Fri, 01 Nov 2002 20:03:49 GMT

Hi Liang,

As I understand, db_block_size specifies the default data block size, it should be multiple of OS level block size. I don't think there's any kernel parameters related to this setting, on HPUX 11, the data block size is 4K, so in my oracle instance, db_block_size=8096.

Regards,
Gary

Re: db_block_size

Bill Thorsteinson — Mon, 04 Nov 2002 16:36:31 GMT

The UNIX block size is set during partioning the disk. Normally this ranges from 512 bytes to 8k. If you are
dedicating the disk to the database then you can set
the block size higher.

The critical calculation is
db_block_size * multiblock_count <= UNIX_block size
Some write intensive systems may perform better with
db_block_size <= UNIX_block_size
All sizes are powers of 2.

This is almost always the case. Changing db_block_size
has requires rebuilding the database. Changing UNIX block size required rebuilding the partition. Multiblock_count is a startup pararameter and can be changed on each database startup.

Don't forget to set the minumum extent when you create
your tablespaces. It should be a power of 2 > blocksize.

Re: db_block_size

Bill Thorsteinson — Mon, 04 Nov 2002 16:43:42 GMT

I flipped first the comparison. It should read.

db_block_size * multi_block_count >= UNIX block size

Equal should mimimize the system reads per database multi-block read.

If db_block_size is greater than UNIX block size you may get multiple system level io requests for all database accesses.