Re: Add min and max execute statement time in pg_stat_statement - Mailing list pgsql-hackers

From KONDO Mitsumasa
Subject Re: Add min and max execute statement time in pg_stat_statement
Date
Msg-id 52FB18A2.20101@lab.ntt.co.jp
Whole thread Raw
In response to Re: Add min and max execute statement time in pg_stat_statement  (KONDO Mitsumasa <kondo.mitsumasa@lab.ntt.co.jp>)
Responses Re: Add min and max execute statement time in pg_stat_statement  (Rajeev rastogi <rajeev.rastogi@huawei.com>)
Re: Add min and max execute statement time in pg_stat_statement  (KONDO Mitsumasa <kondo.mitsumasa@lab.ntt.co.jp>)
List pgsql-hackers
Hi Rajeev,

> (2014/01/29 17:31), Rajeev rastogi wrote:
>> No Issue, you can share me the test cases, I will take the performance report.
Attached patch is supported to latest pg_stat_statements. It includes min, max,
and stdev statistics. Could you run compiling test on your windows enviroments?
I think compiling error was fixed.

We had disscuttion about which is needed useful statistics in community, I think
both of statistics have storong and weak point. When we see the less(2 or 3)
executed statement, stdev will be meaningless because it cannot calculate
estimated value precisely very much, however in this situation, min and max will
be propety work well because it isn't estimated value but fact value. On the
other hand, when we see the more frequency executed statement, they will be
contrary position
statistics, stdev will be very useful statistics for estimating whole statements,
and min and max might be extremely value.
At the end of the day, these value were needed each other for more useful
statistics when we want to see several actual statments. And past my experience
showed no performance problems in this patch. So I'd like to implements all these
values in pg_stat_statements.

Regards,
--
Mitsumasa KONDO
NTT Open Source Software Center






Attachment

pgsql-hackers by date:

Previous
From: "Inoue, Hiroshi"
Date:
Subject: Re: narwhal and PGDLLIMPORT
Next
From: Kyotaro HORIGUCHI
Date:
Subject: Re: Retain dynamic shared memory segments for postmaster lifetime