Thread: How Many Partitions are Good Performing

 You should have read carefully what I wrote.  1000 is not an upper limit.  1000 partition is the number after which
performancestarts dropping .   

There is a blog in www.timescale.com which also highlights the same.

Sent: Monday, January 08, 2018 at 6:20 PM
From: "Kumar, Virendra" <Virendra.Kumar@guycarp.com>
To: "pgsql-general@postgresql.org" <pgsql-general@postgresql.org>
Subject: How Many Partitions are Good Performing

Can somebody tell us how many partitions are good number without impacting the performance. We are hearing around a
thousand,is that a limit. Do we have plan to increase the number of partitions for a table. We would appreciate if
somebodycan help us with this? 
 
Regards,
Virendra
 
 
------------------------------------------------------------
This message is intended only for the use of the addressee and may contain
information that is PRIVILEGED AND CONFIDENTIAL.

If you are not the intended recipient, you are hereby notified that any
dissemination of this communication is strictly prohibited. If you have
received this communication in error, please erase all copies of the message
and its attachments and notify the sender immediately. Thank you.

On Tue, Jan 09, 2018 at 12:54:18AM +0100, Rakesh Kumar wrote:
> Can somebody tell us how many partitions are good number without
> impacting the performance. We are hearing around a thousand, is that a
> limit. Do we have plan to increase the number of partitions for a
> table. We would appreciate if somebody can help us with this?

What matters here is that the bottleneck comes from the planner which
uses basically a O(N^2) algorithm to evaluate all the partitions, so a
too high number causes planning time to increase dramatically. In
Postgres 11, things get improved with more partition-wise logics.
--
Michael

I've run once a test on my laptop because was curious as well. From my
results (on laptop - 16GB RAM, 4 cores) the upper limit was 12k. Above it
planning time was unbearable high - much higher than execution time. It's
been tested on 9.5



--
Sent from: http://www.postgresql-archive.org/PostgreSQL-general-f1843780.html

For test I created two tables with 7800 partitions each and joining them sees performance bottleneck. It is taking 5
secondsplanning time. Please see attached plan. 


Regards,
Virendra

-----Original Message-----
From: pinker [mailto:pinker@onet.eu]
Sent: Wednesday, January 10, 2018 12:07 PM
To: pgsql-general@postgresql.org
Subject: Re: How Many Partitions are Good Performing

I've run once a test on my laptop because was curious as well. From my results (on laptop - 16GB RAM, 4 cores) the
upperlimit was 12k. Above it planning time was unbearable high - much higher than execution time. It's been tested on
9.5



--
Sent from: http://www.postgresql-archive.org/PostgreSQL-general-f1843780.html



________________________________

This message is intended only for the use of the addressee and may contain
information that is PRIVILEGED AND CONFIDENTIAL.

If you are not the intended recipient, you are hereby notified that any
dissemination of this communication is strictly prohibited. If you have
received this communication in error, please erase all copies of the message
and its attachments and notify the sender immediately. Thank you.

yes, it doesn't look good. and it seems that statistics aren't accurate:
GroupAggregate  (cost=271794.39..330553.67 rows=215630 width=152) (actual
time=30.641..37.303 rows=2792 loops=1)



--
Sent from: http://www.postgresql-archive.org/PostgreSQL-general-f1843780.html