On Wed, Jun 26, 2019 at 8:05 AM John Lumby <johnlumby@hotmail.com> wrote:
> > There is a very recent research paper that discusses the idea of
> > varying fillfactor with a view to ameliorating page splits:
> >
>
> Thanks, I am chewing my way through that. As you say, it does address exactly the issues I raised.
> Do you happen to know if their source-code is available somewhere?
> ( I did see their db is MS SQL Server but it still might provide some portable ideas. )
It's just a research paper. It might never be implemented in any system.
The point of the paper is to make page splits occur at a steady rate
after REINDEX'ing -- not to eliminate or even reduce page splits.
> > I suspect that you might find that the enhancements to B-Tree indexes
> > that went into Postgres 12 would help with this workload, especially
> > if you notice that this happens with indexes that have a lot of duplicates
> >
>
> I had not noticed that, thanks for pointing it out. Yes , in my workload most of the indexes in question are
non-uniqueand some have very low key card. I will try out the pg-12 beta when I get a chance.
It's easy to show problems with very low cardinality indexes in the
old code. You'll definitely notice a difference there.
> Is there a pdf or text version?
Just the talk slides:
https://www.pgcon.org/2019/schedule/attachments/518_nbtree-arch-pgcon.pdf
--
Peter Geoghegan