grostoon wrote:

> I'm going to do the same test with postgresql, mysql (with innodb tables) and oracle to see what timings I get.
>

Fine, but do understand that you've chosen the worst possible
set of operations for Firebird. Like most things in life, database
system design is a series of trade-offs. The same multi-version
concurrency control that makes count (*) slow and imposes a clean-up
cost on transactions that follow a large delete also gives us
non-blocking reads, instant recovery after a system crash, and
avoids the need for separate logs, journals, vacuum operations, etc.
If you don't care about multi-user performance, your tests are fine,
but if you do care, test a reasonable mix of concurrent update,
insert, and reporting transactions before you pick a database.

Regards,


Ann