Ann,

> > Zoran's complaint is about the first time the query is executed. In that
> case, nothing would be in cache and everything would still need to be loaded
> from disk.
> >
> A 2048 page cache may be small enough that it would cause problems even
> during loading, given the size of his database and the tables involved. Having
> to reload the same pointer pages would exacerbate his problems. Given that
> the access is indexed, pointer pages and intermediate index pages may need
> to be loaded several times after they've forced all the metadata pages out.

In the latest stats (post 8KB page restore) there are only 248 pages being read from disk, so that would fit into the cache without issues, no?

> It can't hurt to us a much larger cache.

Bigger is always better (except when it comes to male anatomy ;-o)


Sean