Jeremy Palmer <JPalmer(at)linz(dot)govt(dot)nz> wrote:
> My only question is how often will I need to re-cluster the
> table, because it comes at quite a cost.
I probably should have mentioned that the CLUSTER will run faster if
the data is already mostly in the right sequence. You'll be doing a
nearly sequential pass over the heap, which should minimize seek
time, especially if the OS notices the pattern and starts doing
sequential read-ahead.
-Kevin