Re: ToDo list update for BRIN indexes

On 21 December 2015 at 14:38, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:

> On Mon, Dec 21, 2015 at 8:06 AM, Simon Riggs <simon(at)2ndquadrant(dot)com>
> wrote:
> > On 21 December 2015 at 12:54, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> >>
> >> On Thu, Jul 9, 2015 at 4:49 PM, Jeff Janes <jeff(dot)janes(at)gmail(dot)com>
> wrote:
> >> > Is there anything in the below section which has been been implemented
> >> > or
> >> > rendered irrelevant by BRIN indexes?
> >> >
> >> > https://wiki.postgresql.org/wiki/Todo#Indexes
> >> >
> >> > "Consider smaller indexes that record a range of values per heap page,
> >> > rather than having one index entry for every heap row"
> >>
> >> [ catching up on old threads ]
> >>
> >> BRIN is exactly this, isn't it? Well, moreso: it's a range of values
> >> for a range of heap pages.
> >
> > It's close, but not the same.
> >
> > BRIN is a summary index and so could never support uniqueness.
>
> Hmm, but that Todo wording seems to suggest a summary index, so I
> don't think that proposal would support uniqueness either.

Probably garbled slightly.

> > It's also possible to have an index type that has a precise TID entry,
> yet a
> > more compact format, which would then allow unique values. This would be
> > similar to the way SQLServer compresses primary key indexes.
>
> True. But would that require a new index type, or would we do that
> just by optimizing btree?

Heikki worked on Grouped Item Tuple indexes about 8 years ago doing exactly
that. They gave about 30% performance advantage at the time, when used with
HOT, which was the original driver for that feature.

BRIN and GIT approaches degrade under heavy non-HOT updates.

--
Simon Riggs http://www.2ndQuadrant.com/
<http://www.2ndquadrant.com/>
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services