Re: Locking B-tree leafs immediately in exclusive mode

Hi!

On Wed, Jul 25, 2018 at 1:19 PM Simon Riggs <simon(at)2ndquadrant(dot)com> wrote:
> On 13 July 2018 at 03:14, Imai, Yoshikazu <imai(dot)yoshikazu(at)jp(dot)fujitsu(dot)com> wrote:
> > From an attached graph("some_contention_points_on_leaf_nodes.png"), as contention points dispersed, we can see that TPS is increased and TPS difference between master and patched version becomes smaller.
>
> So I think this clearly shows the drop in throughput when we have
> index contention and that this patch improves on that situation.
>
> In cases where we don't have contention, the patch doesn't cause a
> negative effect.
>
> So +1 from me!

Thank you, but Imai found another case [1], where patch causes a small
regression. For now, it's not completely clear for me what test case
was used, because I forgot to add any indexes to initial specification
of this test case [2]. If regression will be confirmed, then it would
be nice to mitigate that. And even if we wouldn't manage to mitigate
that and consider that as acceptable trade-off, then it would be nice
to at least explain it...

1. https://www.postgresql.org/message-id/0F97FA9ABBDBE54F91744A9B37151A51189451%40g01jpexmbkw24
2. https://www.postgresql.org/message-id/CAPpHfds%3DWmjv%2Bu7S0peHN2zRrw4C%3DYySn-ZKddp4E7q8KQ18hQ%40mail.gmail.com------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company