Re: 9i multi cache buffer

14 Nov 2002 04:38:56 -0800, Geoff Ingram said (and I quote):
>
> Oracle announced a record breaking 4CPU TCP-C (Linux and HPUX,
> identical Intel hardware, Oracle10) to coincide with OracleWorld. This

Ay! TPC-C benchmarks are worth virtually nothing....

> was achieved using multiple buffer caches:
>
> db_cache_size = 3500M
> db_8k_cache_size = 512M
> db_16k_cache_size = 3000M
> db_keep_cache_size = 37750M

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

I beg your pardon??????

> db_recycle_cache_size = 50M
>
> TPC-C is an extreme OLTP workload characterised by small transactions
> and concurrent block access by multiple sessions. In this case a small
> blocksize is good - evidenced by the fact the TPC-C database uses
> db_block_size=2048.

Yes. It probably used nothing but the small block size cache. That's what TPC-C does. None of the others would have been used...

> Same for UNDO, for which Oracle used 8K blocks in the benchmark.

Strange...

> What about a situation where your table has large rows (say 16K each)
> with small index entries (say a numeric primary key)? Do you really
> want them both stored in blocks of the same size?

Yes. Most of the time.

>
> Multiple buffer caches can significantly improve performance in the
> right circumstances. Only you can tell whether they apply to your
> situation.

I'm not sure the above proves anything re performance improvements by using multiple block sizes in multiple buffer caches.

-- 
Cheers
Nuno Souto
nsouto_at_optusnet.com.au.nospam