Re: Postgres not using all RAM (Huge Page activated on a96GB RAM system) - Mailing list pgsql-performance

From Shaun Thomas
Subject Re: Postgres not using all RAM (Huge Page activated on a96GB RAM system)
Date
Msg-id CAB78C+CfLgTLWCDukpLo=gA-LWPhJQKAdJhnqAAmhBMBnxP7wg@mail.gmail.com
Whole thread Raw
In response to Re: Postgres not using all RAM (Huge Page activated on a96GB RAM system)  (Álvaro Hernández Tortosa <aht@8kdata.com>)
List pgsql-performance
> What’s wrong with this? There isn’t something wrong in RAM usage?

Nope, nothing wrong with RAM usage at all from what you've presented
here. Please consider the cut-and-paste you included a bit closer. All
of your active threads are utilizing 100% CPU, and are therefore
CPU-bound. If there were some kind of IO issue due to disk fetching,
your CPU utilization would be much lower. From the looks of things,
your threads are either operating on fully cached or otherwise
available pages, or are generating their own such that it doesn't
matter.

The real question is this: what are your queries/processes doing?
Because if the query plan is using a giant nested loop, or you are
relying on a stored procedure that's in a tight and non-optimized loop
of some kind, you're going to be consuming a lot of clock cycles with
diminishing benefits. If you're not making use of set theory within a
database, for example, you might be getting 100x less throughput than
you could otherwise attain. If it's not proprietary in some way, or
you can obfuscate it into a test case, we can probably help then. As
it stands, there isn't enough to go on.

--
Shaun Thomas
bonesmoses@gmail.com
http://bonesmoses.org/


pgsql-performance by date:

Previous
From: Álvaro Hernández Tortosa
Date:
Subject: Re: Postgres not using all RAM (Huge Page activated on a96GB RAM system)
Next
From: trafdev
Date:
Subject: Re: Postgres not using all RAM (Huge Page activated on a96GB RAM system)