Nico Callewaert wrote:

>
> Hi again,
>
> I don't know if I have to disagree with Paul, but as a test I backuped and
> restored the database and I changed the buffersize from 4096 to 2048 and it
> looks slower to me now ???
>

Could you define 'looks slower' - is it a new term that I have missed? <g>.

There are two problems in this area. One is the complete lack of research, or
even the specification of the testing required.

The other is the number of factors to consider. Inevitably most tests will be
done in an unreal 'laboratory' environment. That aside, InterBase will behave
differently on different OS's, with different physical RAM, differing disc
sub-systems, network infrastructure, application design, database usage and
number of concurrent users.

All I found was that with the particular tests I did I produced particular
results. Hopefully I can prepare something soon that would enable tests to be
run on more systems, so that we can collate the results and get a better
understanding of what is happening.

But in the meantime, could you try some unscientific tests of your own, via
WISQL or similar, with the statistics turned on - that way you will be able to
compare execution times and see the number of fetches, reads etc. Make sure that
you execute every select statement about three times. (And record the stats for
each - they tell a story.)

A quicker way of changing the buffer size is with gfix - that way you can skip
the backup and restore cycle. (Disconnect first, though).


Paul
--

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