Re: Date vs Timestamp without timezone Partition Key - Mailing list pgsql-performance

From Cedric Leong
Subject Re: Date vs Timestamp without timezone Partition Key
Date
Msg-id CAD6i=X1RWpg1GR0-h-KO0n=WE6PKoWeUs-pytgX6sfZSczQ54w@mail.gmail.com
Whole thread Raw
In response to Re: Date vs Timestamp without timezone Partition Key  (Tom Lane <tgl@sss.pgh.pa.us>)
Responses Re: Date vs Timestamp without timezone Partition Key  (David Rowley <dgrowleyml@gmail.com>)
List pgsql-performance
Somewhat unrelated but note to anyone who wants to swap out partition keys. Don't create a clone of the table with the new partition key and insert data. It messes up the query planner massively and makes everything much slower.

On Mon, May 25, 2020 at 12:48 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
Cedric Leong <cedricleong@gmail.com> writes:
> Just in case someone is interested enough to answer this. Does anyone know
> if the performance for a date column vs a timestamp column as a partition
> key is large?

I doubt it's even measurable, at least on 64-bit machines.  You're
basically talking about 32-bit integer comparisons vs 64-bit integer
comparisons.

On a 32-bit machine it's possible that an index on a date column
will be physically smaller, so you could get some wins from reduced
I/O.  But on (most?) 64-bit machines that difference goes away too,
because of alignment restrictions.

As always, YMMV; it never hurts to do your own testing.

                        regards, tom lane

pgsql-performance by date:

Previous
From: Imre Samu
Date:
Subject: Re: When to use PARTITION BY HASH?
Next
From: David Rowley
Date:
Subject: Re: Date vs Timestamp without timezone Partition Key