On Tue, Apr 17, 2018 at 11:09 AM, Tom Lane <t...@sss.pgh.pa.us> wrote: > So rather than a klugy solution that only fixes > VACUUM (and not very well, requiring user intervention and an unpleasant > tradeoff), we ought to look at ways to avoid needing a whole-pool scan to > find the pages belonging to one relation. In the past we've been able to > skate by without a decent solution for that because shared buffers were > customarily not all that big. But if we're going to start considering > huge buffer pools to be a case we want to have good performance for, > that's got to change.
Andres mentioned that he has prototyped an approach to buffer management that uses a Radix tree, which is generally assumed to be the right long-term fix. -- Peter Geoghegan
