From: Karl Denninger <karl@denninger.net>
To: pgsql-admin@lists.postgresql.org
Subject: Re: How to calculate how too much memory ram need a PostgreSQL Server?
Date: Mon, 5 Apr 2021 14:27:59 -0400
Message-ID: <a8ba4eba-8fdd-92c5-e6e3-f4c734050eb2@denninger.net> (raw)
In-Reply-To: <1617646911570-0.post@n3.nabble.com>
References: <1617646911570-0.post@n3.nabble.com>
On 4/5/2021 14:21, DrakoRod wrote:
> Hi folks!!
>
> I support a PostgreSQL database it grew from 100 to 300 GB in couple months,
> I we need tuning all OS parameters (is a virtual enviroment). Currently
the
> configuration is the follow:
>
> Virtual Machine
> Ubuntu 18.04
> 12 CPUS
> 8 GB in RAM
> 400 in Disk
> Clúster database size 320 GB
>
> Some queries have times too differents, the same query sometimes it execute
> in 8 10 seconds sometimes in 12 15 minutes, many querys make joins between
> 40, 50 millions rows. I know some factors like relation between entities
> affect times, but in general I suspect the mainly problem si resources in
> OS.
>
> I suspect is a lack to memory ram but we need justify the change in OS RAM,
> so, How to calculate how too much memory ram need a PostgreSQL Server? Too
> need know how because the database size continue grows.
>
> Best Regards
> DrakoRod
IMHO you need to profile the workload (e.g. "Explain analyze" on the
queries in question), look at the indices in use and whether the planner
can be helped with additional ones on the tables and then look at the
size of said indices and how all of this fits into the RAM you have and
what storage is behind it ( RAM -> SSD -> spinning rust.)
When the working set of the indices required for the queries in question
(not so much the data itself) get outside of the available RAM space
performance gets hit fast since any sort of storage is a lot slower,
with spinning rust being excruciatingly slower since it has seek time
penalties that SSDs do not.
--
Karl Denninger
karl@denninger.net <mailto:karl@denninger.net>
/The Market Ticker/
/[S/MIME encrypted email preferred]/
Attachments:
smime.p7s (application/pkcs7-signature, 4.8K) download
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-admin@postgresql.org
Cc: karl@denninger.net, pgsql-admin@lists.postgresql.org
Subject: Re: How to calculate how too much memory ram need a PostgreSQL Server?
In-Reply-To: <a8ba4eba-8fdd-92c5-e6e3-f4c734050eb2@denninger.net>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox