Re: POC: Cache data in GetSnapshotData()

On Thu, Sep 21, 2017 at 7:25 AM, Robert Haas <robertmhaas(at)gmail(dot)com> wrote:
> On Wed, Sep 20, 2017 at 9:46 PM, Mithun Cy <mithun(dot)cy(at)enterprisedb(dot)com> wrote:
>> I think there is some confusion above results is for pgbench simple update
>> (-N) tests where cached snapshot gets invalidated, I have run this to check
>> if there is any regression due to frequent cache invalidation and did not
>> find any. The target test for the above patch is read-only case [1] where we
>> can see the performance improvement as high as 39% (@256 threads) on
>> Cthulhu(a 8 socket numa machine with 64 CPU cores). At 64 threads ( = CPU
>> cores) we have 5% improvement and at clients 128 = (2 * CPU cores =
>> hyperthreads) we have 17% improvement.
>>
>> Clients BASE CODE With patch %Imp
>>
>> 64 452475.929144 476195.952736 5.2422730281
>>
>> 128 556207.727932 653256.029012 17.4482115595
>>
>> 256 494336.282804 691614.000463 39.9075941867
>
> Oh, you're right. I was confused.
>
> But now I'm confused about something else: if you're seeing a clear
> gain at higher-client counts, why is Jesper Pederson not seeing the
> same thing? Do you have results for a 2-socket machine? Maybe this
> only helps with >2 sockets.

My current tests show on scylla (2 socket machine with 28 CPU core) I
do not see any improvement at all as similar to Jesper. But same tests
on power2 (4 sockets) and Cthulhu(8 socket machine) we can see
improvements.

--
Thanks and Regards
Mithun C Y
EnterpriseDB: http://www.enterprisedb.com