[pgpool-hackers: 3228] Re: pgbench: Flamegraph

[Tatsuo Ishii]

> Hi,
> 
> I did a flamegraph of v3.7.x running
> 
>  pgbench -h localhost -p 9999 -U pgbench -c 8 -j 8 -T 60 pgbench
> 
> Going direct (-p 5432) gave 9738 tps, where as pgpool gave 3284
> tps. One thing to note is that the entire stack was on one machine.
> 
> The pgpool configuration was done for a single instance against a
> single instance of PostgreSQL 10.x.
> 
> If there is an interest in this I can setup the environment on
> separate machines, and create the profile using master.

Thanks for the flamegraph. I am new to the tool but it looks pretty
interesting. Is it possible for you to take another flamegraph with
non extended query case (the pgbench default or "-M simple") so that
I could get a hint to find bottle comparing -M extended case and -M
simple case. Yes, we already noticed the performance degradation with
Pgpool-II but still are struggling how to mitigate it.

Also it would be nice if you kindly show us how to generate flamegraph
of Pgpool-II.

Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp