> On Mon, 13 Aug 2018 at 17:36, Alexander Korotkov <a.korotkov@postgrespro.ru> wrote:
>
> Hi!
>
> I run pgbench (read-write and read-only benchmarks) on Amazon
> c5d.18xlarge virtual machine, which has 72 VCPU (approximately same
> power as 36 physical cores). The results are attached
> (lwlock-fair-ro.png and lwlock-fair-rw.png).
Hi,
Thanks for working on that. I haven't read the patch yet, but I think it worth
mentioning that with testing locks on AWS we also need to take into account
lock holder/waiter preemtion problem and such things as Intel PLE, since I
believe they can have significant impact in this case.
Right now I'm doing some small research on this topic and I hope soon I'll
finish a tool and benchmark setup to test the influence of such factors in
virtualized environment. In the meantime I would be glad to review your
patches.