Subject: Re: Make autovacuum sort tables in descending order of xid_age
From: Mark Dilger <hornschnorter@gmail.com>
To: David Fetter <david@fetter.org>
Cc: PostgreSQL Development <pgsql-hackers@postgresql.org>,
 Christophe Pettus <xof@thebuild.com>
References: <20191129180138.GD8731@fetter.org>
 <20191129222152.GE8731@fetter.org>
 <2488acea-328b-1328-d437-8c156bc8f7a4@gmail.com>
 <20191130222314.GF8731@fetter.org>
 <877ac074-e95d-bdee-09e7-c51f0ee0fde2@gmail.com>
 <20191212192631.GA29201@fetter.org>
 <2ad2a9fa-32eb-29aa-07ee-f0fe75ad4db5@gmail.com>
Message-ID: <49dac8e3-40bc-2986-e873-51f874c0f5fa@gmail.com>
Date: Thu, 12 Dec 2019 13:47:23 -0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:68.0) Gecko/20100101
 Thunderbird/68.2.2
MIME-Version: 1.0
In-Reply-To: <2ad2a9fa-32eb-29aa-07ee-f0fe75ad4db5@gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk


On 12/12/19 1:35 PM, Mark Dilger wrote:
>    Let C = 1.00000002065
>    Let x = xid_age for a table
>    Let v = clamp(n_dead_tuples / reltuples*2) to max 0.5
>    Let a = clamp(changes_since_analyze / reltuples) to max 0.5
> 
>    Let score = C**x + v + a

I should hasten to add that this is just a proof of concept
formula, not one that I'm specifically advocating.  The point
is that we can devise a scoring system in which the xid age
is the dominant factor whenever it could possibly matter,
while still letting other factors prevail when xid age is
of little consequence.

-- 
Mark Dilger