Our Database is going to be created with a DB_BLOCK_SIZE of 8k on an
Alpha 8400 with RAID 7000 and 0+1 for the level.

Digital says that the UNIX DEV_BSIZ is 512 blocks or 256K.

What is the process for determining the correct DB_BLOCK_SIZE for Oracle
if the Digital says that the UNIX Block size is 512 blocks ?

How does RAID play into determining the DB_BLOCK_SIZE ?

The DB_BLOCK_SIZE should be a multiple of the file system block size, to
avoid reading data that will be discarded. I don't know what DEV_BSIZ, but
256k block size seems an unreasonably high value for a file system. Maybe
what they're really trying to say is that 512 blocks make 256k, hence the
block size is 512 bytes?

One of the most important factors in determining the DB_BLOCK_SIZE is the
type of application. A rule of thumb is to identify whether your application
is mostly an OLTP or mostly an OLAP system. Generally (but not always), OLTP
systems run best with small data blocks, as OLAP systems run best with
larger data blocks.

As for RAID, I think it should play a lesser part in determining the optimal
DB_BLOCK_SIZE.

I think the block size for these Alpha servers is 8k (unless someone knows
otherwise??)

Reply:

I think, that you can set 16K. I saw such a value on DUnix with RAID.

There is a default block size for the filesystem when it is created. Of
course, the default can be changed at creation time.

Have a look at the man pages for 'fstyp' and 'stat'...

Reply:

In Digital UNIX, there is no fstyp command .....