Re: why after increase the hash table partitions, TPMC decrease

On Wed, Sep 3, 2014 at 8:44 AM, Xiaoyulei <xiaoyulei(at)huawei(dot)com> wrote:
>
>
> benchmarSQL has about half reads. So I think it should be effective.
>
> I don't think BufFreelistLock take much time, it just get a buffer from
list. It should be very fast.

Only incase all the data fits in shared buffers, else it needs to
perform clock sweep which can be costly in certain cases.

> The test server has 2 CPUs and 12 cores in each CPU. 24 processor
totally. CPU Idle time is over 50%. IO only 10%(data is in SSD)
>
> I perf one process of pg. The hot spot is hash search. perf data file is
more than 1M, so I do not attach it. I send it separately.

Could you once check the callers of hash_search_with_hash_value()
as it gets called from multiple paths? I am not able to view the
perf.data file you have sent.

Also, you might want to check the performance on 9.4 codebase,
as there are quite a few performance improvements in it.

With Regards,
Amit Kapila.
EnterpriseDB: http://www.enterprisedb.com