Message-ID: <a16ac766-f898-3299-28fd-5aabfa6f308d@enterprisedb.com>
Date: Thu, 18 Aug 2022 09:29:43 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.9.0
Subject: Re: POC: GROUP BY optimization
Content-Language: en-US
To: David Rowley <dgrowleyml@gmail.com>
Cc: Michael Paquier <michael@paquier.xyz>, Zhihong Yu <zyu@yugabyte.com>,
 "Andrey V. Lepikhov" <a.lepikhov@postgrespro.ru>,
 Teodor Sigaev <teodor@sigaev.ru>, Tomas Vondra
 <tomas.vondra@2ndquadrant.com>,
 PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>
References: <721b3ff9-f214-f5d4-86c7-bae515bbec18@enterprisedb.com>
 <7bea462e-8c3e-0f8c-c7d5-f4daa12f3baf@postgrespro.ru>
 <9af89543-3d17-4c98-59de-1daa5bceeb06@enterprisedb.com>
 <9408e450-60c6-6fbc-d5c4-467bb0abfe67@postgrespro.ru>
 <44472925-f6c6-e8de-6f72-0451458532c4@enterprisedb.com>
 <CALNJ-vQp3WNV-kKuGsjSOhjRfdECA1Dbxa-vjAgOmQ4f4Qaw-Q@mail.gmail.com>
 <7a80a207-5e8e-b80c-476c-2d290b0526e9@enterprisedb.com>
 <65941e87-86c1-c09c-8ad5-8e102aebe8ee@enterprisedb.com>
 <CAApHDvpVFgWzXdtUQkjyOPhNrNvumRi_=ftgS79KeAZ92tnHKQ@mail.gmail.com>
 <aada8f97-924e-5661-aead-257aa346899c@enterprisedb.com>
 <Yuj6mQomSuhoti9D@paquier.xyz>
 <CAApHDvruRZ0cVw0rsjoGSgoQF5fn2UoJ2cWLgZSVza0M1Y1KPg@mail.gmail.com>
 <683635b8-381b-5b08-6069-d6a45de19a12@enterprisedb.com>
 <CAApHDvr+95wJcABsqSXWcBa+yyUmJ2bq_mnZs9dUnaOi3XRVMg@mail.gmail.com>
From: Tomas Vondra <tomas.vondra@enterprisedb.com>
In-Reply-To: 
 <CAApHDvr+95wJcABsqSXWcBa+yyUmJ2bq_mnZs9dUnaOi3XRVMg@mail.gmail.com>
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit
Archived-At: 
 <https://www.postgresql.org/message-id/a16ac766-f898-3299-28fd-5aabfa6f308d%40enterprisedb.com>
Precedence: bulk


On 8/18/22 03:32, David Rowley wrote:
> On Thu, 18 Aug 2022 at 02:46, Tomas Vondra
> <tomas.vondra@enterprisedb.com> wrote:
>> So I don't think the current costing is wrong, but it certainly is more
>> complex. But the test does not test what it intended - I have two ideas
>> how to make it work:
>>
>> 1) increase the number of rows in the table
>>
>> 2) increase cpu_operator_cost (for that one test?)
>>
>> 3) tweak the costing somehow, to increase the cost a bit
> 
> Why not, 4) SET parallel_setup_cost = 0;  there are plenty of other
> places we do just that so we get a parallel plan without having to
> generate enough cost to drown out the parallel worker startup cost.
> 
> Here are a couple of patches to demo the idea.
> 

Yeah, that's an option too. I should have mentioned it along with the
cpu_operator_cost.

BTW would you mind taking a look at the costing? I think it's fine, but
it would be good if someone not involved in the patch takes a look.


regards

-- 
Tomas Vondra
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company