Hi,
> it's a lot of work and right now the only people
> who've done that work aren't giving it away for free - or not in any form
> that can be integrated into PostgreSQL without removing other capabilities
> other users need.
One MPP vendor implemented columnar store in roughly six months --
lot's of work is involved there!. Anyway, all implementation what I
came across took several shortcuts like no updates(append only) or no
foreign keys, ... but it works!
> That's not to say Pg can't improve. It can, and not just by adding column
> store or index-structured table support. Improved parallelism capabilities
> are needed in Pg
I see most benefits coming from parallelism: 12hr query can finish in
2hr if sliced properly
--
Ondrej Ivanic
(ondrej.ivanic@gmail.com)