pwing@xxxxxxxxxxxxxxxxx writes: > I am running three ways: sequential scan, bitmap index scan and index scan. > The I/O cost for the index scan is 24+ times more than the other two. I do not > understand why this happens. If I am using a clustered index, it is my > understanding that there should be no need to retrieve a single page multiple > times, as tuples are sorted. Am I misunderstanding something? How long ago did you cluster the table? It sounds like it has degraded far away from a fully-ordered state. regards, tom lane
