Re: Move PinBuffer and UnpinBuffer to atomics

On Sat, Apr 9, 2016 at 11:24 AM, Alexander Korotkov <
a(dot)korotkov(at)postgrespro(dot)ru> wrote:

> On Fri, Apr 8, 2016 at 10:19 PM, Alexander Korotkov <
> a(dot)korotkov(at)postgrespro(dot)ru> wrote:
>
>> On Fri, Apr 8, 2016 at 7:39 PM, Andres Freund <andres(at)anarazel(dot)de> wrote:
>>
>>> As you can see in
>>>
>>
>>> http://archives.postgresql.org/message-id/CA%2BTgmoaeRbN%3DZ4oWENLvgGLeHEvGZ_S_Z3KGrdScyKiSvNt3oA%40mail.gmail.com
>>> I'm planning to apply this sometime this weekend, after running some
>>> tests and going over the patch again.
>>>
>>> Any chance you could have a look over this?
>>>
>>
>> I took a look at this. Changes you made look good for me.
>> I also run test on 4x18 Intel server.
>>
>
> On the top of current master results are following:
>
> clients TPS
> 1 12562
> 2 25604
> 4 52661
> 8 103209
> 10 128599
> 20 256872
> 30 365718
> 40 432749
> 50 513528
> 60 684943
> 70 696050
> 80 923350
> 90 1119776
> 100 1208027
> 110 1229429
> 120 1163356
> 130 1107924
> 140 1084344
> 150 1014064
> 160 961730
> 170 980743
> 180 968419
>
> The results are quite discouraging because previously we had about 1.5M
> TPS in the peak while we have only about 1.2M now. I found that it's not
> related to the changes you made in the patch, but it's related to 5364b357
> "Increase maximum number of clog buffers". I'm making same benchmark with
> 5364b357 reverted.
>

There are results with 5364b357 reverted.

clients TPS
1 12980
2 27105
4 51969
8 105507
10 132811
20 256888
30 368573
40 467605
50 544231
60 590898
70 799094
80 967569
90 1211662
100 1352427
110 1432561
120 1480324
130 1486624
140 1492092
150 1461681
160 1426733
170 1409081
180 1366199

It's much closer to what we had before.

------
Alexander Korotkov
Postgres Professional: http://www.postgrespro.com
The Russian Postgres Company