> I think you people are missing the point here, the performance increase > you're seeing is all down to OS caching and will vary across different > ports. It's nothing to do with sqlite, and will affect every package. > > Therefore the only way to fairly compare mysql/postgress/sqlite is to make > sure the machine is cleanly booted, before running any tests. > > ( well and then maybe run the tests twice in succession, so the caching > effect can be taken into account )
This was what we found, and we tested dozens of databases and home grown systems Some tools did better with initial cacheing and some better once cached All our speed tests started on a clean boot, then we tested several iterations. Even delphi wrappers for the same back end (in one case SQLite) differed in the speed they worked at. Has taken months to get to this point where we are finally happy with the results we are getting from a DB