[pgpool-hackers: 732] Re: performance degrade?

[Muhammad Usama]

Hi

I am getting almost identical performance with pgpool-II 3.3 and 3.4
by using the current git head for both the branches.
I can think of one memory leak, checked-in on December 3rd 2014
(commit: 2637fb3a18b9422178ce0b5f36bd99fec852f728), that can cause the
degration in the performance, If the 3.4 installer used for
benchmarking does not containes that fix.


3.4 BRANCH
==========
bin/pgpool --version
pgpool-II version 3.4.0 (tataraboshi)

bin$ ./pgbench -p 9999 -c 10 -T 300 bench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 1
query mode: simple
number of clients: 10
number of threads: 1
duration: 300 s
number of transactions actually processed: 270639
latency average: 11.085 ms
tps = 902.106169 (including connections establishing)
tps = 902.348019 (excluding connections establishing)


3.3 BRANCH
=============
bin/pgpool --version
pgpool-II version 3.3.4 (tokakiboshi)

/bin$ ./pgbench -p 9999 -c 10 -T 300 bench
starting vacuum...end.
transaction type: TPC-B (sort of)
scaling factor: 1
query mode: simple
number of clients: 10
number of threads: 1
duration: 300 s
number of transactions actually processed: 270799
latency average: 11.078 ms
tps = 902.639811 (including connections establishing)
tps = 902.699808 (excluding connections establishing)

Kind regards
Muhammad Usama

On Mon, Dec 22, 2014 at 11:17 AM, Tatsuo Ishii <ishii at postgresql.org> wrote:
> Usama,
>
> Any comment on this?
>
> Best regards,
> --
> Tatsuo Ishii
> SRA OSS, Inc. Japan
> English: http://www.sraoss.co.jp/index_en.php
> Japanese:http://www.sraoss.co.jp
>
>> Hello
>>
>> I'm trying simple performance test of pgpool using pgbench.
>> Then I noticed pgpool-3.4.0 has lower performance than 3.3.4.
>>
>> I've changed some parameter in pgpool.conf a little
>> for adjusting to my environment.
>>
>> --
>> listen_addresses = '*'
>> backend_port0 = 22934
>> backend_data_directory0 = '/home/katsumata/local/pgsql934/data
>> num_init_children = 20
>> --
>>
>> Other parameters are used with default value in pgpool.conf.
>>
>> The results are below.
>>
>> ------------------------------------------------------------
>> 3.4.0
>> $ pgbench -p 9999 -c 10 -T 300 bench
>> starting vacuum...end.
>> transaction type: TPC-B (sort of)
>> scaling factor: 1
>> query mode: simple
>> number of clients: 10
>> number of threads: 1
>> duration: 300 s
>> number of transactions actually processed: 77189
>> tps = 256.613427 (including connections establishing)
>> tps = 256.701916 (excluding connections establishing)
>>
>> ------------------------------------------------------------
>> 3.3.4
>> $ pgbench -p 9999 -c 10 -T 300 bench
>> starting vacuum...end.
>> transaction type: TPC-B (sort of)
>> scaling factor: 1
>> query mode: simple
>> number of clients: 10
>> number of threads: 1
>> duration: 300 s
>> number of transactions actually processed: 109285
>> tps = 364.264430 (including connections establishing)
>> tps = 364.515379 (excluding connections establishing)
>> ------------------------------------------------------------
>>
>> I set checkpoint_segments to 100, and before running
>> pgbench I execute CHECKPOINT to reduce its impact.
>>
>> Am I missing some settings for good performance?
>>
>> Another Information:
>> I'm using RPM packages, and the versions are below.
>> - pgpool-II-pg93-3.4.0-1.pgdg.x86_64.rpm
>> - pgpool-II-pg93-3.3.4-2.pgdg.x86_64.rpm
>>
>> regards,
>> ----------------
>> Tomonari Katsumata
>>
>>
>> _______________________________________________
>> pgpool-hackers mailing list
>> pgpool-hackers at pgpool.net
>> http://www.pgpool.net/mailman/listinfo/pgpool-hackers
> _______________________________________________
> pgpool-hackers mailing list
> pgpool-hackers at pgpool.net
> http://www.pgpool.net/mailman/listinfo/pgpool-hackers