>for example for a webapp a 0.25 second query that's run on every pageload is extremely slow, and creating an index for that is almost certainly a good trade-of

>for example for a webapp a 0.25 second query that's run on every pageload is extremely slow, and creating an index for that is almost certainly a good trade-off. On the hand, it's usually fine if some analytical query that only your finance people need once a month runs for a minute; in that case, an index is probably a poor trade-off. But that should be automatable. If the DB sees frequent queries, it would create indexes, if those queries cease, it would drop them. If queries are infrequent, it would not create an index.