Re: Re: lwlocks and starvation

Neil Conway <neilc(at)samurai(dot)com> wrote on 02.12.2004, 05:55:43:
> On Wed, 2004-12-01 at 21:51 -0500, Bruce Momjian wrote:
> > Neil, where are we on this? Should we add comments? Add a TODO? A patch?
>
> I'm not sure what the right resolution is. As I said, I don't think it's
> wise to apply a patch that could have a significant impact on
> performance without (a) testing its performance effect and/or (b) having
> any evidence that the problem it addresses actually effects anyone in
> the real world.

I agree, it would be unwise to apply the patch. In contrast to my
comments on the nodeAgg perf tweak, changing the way lwlocks are
allocated could have catastrophic effect on the whole system. We've got
no evidence that it repesents a beneficial performance change, but the
risk of it causing some unforeseen problem in at least one corner of
the code is seems very high.

I'm pleased we found this, but changing it should be deferred till at
least 8.1. Perhaps a TODO item could read "investigating lwlock
scheduling algorithms"? We might yet find that there is no single right
answer and we need different algorithms for different locks...

Best Regards, Simon Riggs