DBWn tuning does not seem to be as much of an issue at 8.1 as before. At least,
I personally have not yet had to contend with any big DBWn tuning problems at
8.1. It is the "DBWR write chunk" (kcbswc not kcbscw) that corresponds to the
old "DB writer IO clump". The value is now set with _db_writer_chunk_writes
rather than _db_block_write_batch. The default value appears to be a platform
specific constant - that is, not derived from any other parameters.

The "new" parameter is a limit on the maximum number of outstanding I/O
operations for all DBWn processes. It can be set with _db_writer_max_writes.
Previously this was limited by the Oracle kernel constant SSTMXIOB, and
customers had to tune down the write batch size to throttle down the number of
simultaneous asynchronous writes if necessary (some operating systems are prone
to losing asynchronous writes if there are too many concurrent requests).


-----Original Message-----
Sent: Tuesday, 29 August 2000 20:14


In Oracle 8.0 my X$KVII shows the DBWR statistic as

DB writer IO clump              50  (#buffers/4)
DB writer checkpoint clump      16

In 8.1 X$KVII new statistic shows as

DBWR max outstanding writes    4096  (#buffers/2)
DBWR write chunk               204

Can  you tell about the new stat and is it because of the new DBWR algorithm? I
have Oracle 8 buffer cache management whitepaper (from Oracle Center of
Excellence). By anyway can i get the same for Oracle 8.1?