> You might have forgotten to shutdown and restart the server between
> both test-queries. Else the second run would have significantly less
> disk IO due to the database cache.
>

The database cache seems not to be used at all. The only difference I can see in Filemons log is that there are failures to read the data from Windows system cache. Filemon output looks like this:

954 15:51:06.392 fbserver.exe:1360 IRP_MJ_READ* C:\LOKAL\TST-FB\DATA\NORMRUN.FDB SUCCESS Offset: 73736192 Length: 4096
955 15:51:06.392 fbserver.exe:1360 FASTIO_READ C:\LOKAL\TST-FB\DATA\NORMRUN.FDB SUCCESS Offset: 73736192 Length: 4096
....
upto 20'000 disk I/Os for my select.
See other postings for more info.


Guenther