> From: Peter Geoghegan <pg@xxxxxxx> > Sent: July 8, 2019 1:39 PM > Subject: Re: REINDEX : new parameter to preserve current average leaf density as new implicit FILLFACTOR > > On Mon, Jul 8, 2019 at 9:23 AM John Lumby <johnlumby@xxxxxxxxxxx> wrote: > > > Although a welcome improvement, I think it is not enough to justify stopping use of setting > > a lower explicit FILLFACTOR. Which then brings me back to thinking there is a case > > for the subject of this thread, an automatic way to preserve density. > > I don't think that such an option would make much sense. The "waves of > misery" paper is about smoothing out the frequency of page splits > following bulk loading and a CREATE INDEX. It is not about making > splits occur less often. It's well understood that a certain amount of > free space is the overhead of B-Tree indexes, albeit an overhead that > can be avoided in certain specific instances. > Yes, I see that. But surely "making splits occur less often" is a desirable objective in itself, is it not? And I believe that a parameter to preserve the "steady-state" density in high-traffic indexes would help achieve that goal, wouldn't you agree? Cheers, John Lumby
