Date: Mon, 28 Nov 2022 21:34:58 -0600
From: Justin Pryzby <pryzby@telsasoft.com>
To: Josh Berkus <josh@berkus.org>
Cc: pgsql-performance@postgresql.org
Subject: Re: Catching up with performance & PostgreSQL 15
Message-ID: <20221129033458.GF11463@telsasoft.com>
References: <89e7da59-0dfe-45fe-e082-00f777e200a1@berkus.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <89e7da59-0dfe-45fe-e082-00f777e200a1@berkus.org>
User-Agent: Mutt/1.9.4 (2018-02-28)
Archived-At: 
 <https://www.postgresql.org/message-id/20221129033458.GF11463%40telsasoft.com>
Precedence: bulk

On Mon, Nov 28, 2022 at 06:59:41PM -0800, Josh Berkus wrote:
> Hey, folks:
> 
> I haven't configured a PostgreSQL server since version 11 (before that, I
> did quite a few).
> 
> What's changed in terms of performance configuration since then?  Have the
> fundamentals of shared_buffers/work_mem/max_connections changed at all?
> Which new settings are must-tunes?
> 
> I've heard about new parallel stuff an JIT, but neither is that applicable
> to my use-case.

shared buffers is the same, but btree indexes are frequently (IME) 3x
smaller (!) since deduplication was added in v13, so s_b might not need
to be as large.

In addition to setting work_mem, you can also (since v13) set
hash_mem_multiplier.

default_toast_compression = lz4 # v14
recovery_init_sync_method = syncfs # v14
check_client_connection_interval = ... # v14
wal_compression = {lz4,zstd} # v15

Peeking at my notes, there's also: partitioning, parallel query, brin
indexes, extended statistics, reindex concurrently, ...

... but I don't think anything is radically changed :)

-- 
Justin