Nico Callewaert wrote:

>
> Thanks Paul,
>
> And a page size of 4096, is that ok ?
>

Not on the tests I did (very limited - networked client against an NT box w/
64MB ram.) 4096 consistently came out worse than both lesser and greater values.
I do not understand why, but there is definitely a black hole there.

I was testing by doing a 'group by' on a large table. The tests were repeated to
get an average, then the cache was flushed by doing a 'group by' on another
large table.

Interestingly enough it also appears that if the cache is frequently forced to
flush in that manner the cost of memory mgt of a large buffer outweighs the
benefits provided of easy access to data in memory. Which is why overall I would
recommend 1024 for a starting point and do tests to establish if a better
setting can be found for a particular application.


Paul
--

Paul Reeves
http://www.ibphoenix.com
taking InterBase further