Miles Keaton <mileskeaton(at)gmail(dot)com> writes:
> What surprised me the most is that the subset, even in the original
> database, gave search results MUCH faster than the full table!
The subset table's going to be physically much smaller, so it could just
be that this reflects smaller I/O load. Hard to tell without a lot more
detail about what case you were testing.
regards, tom lane