From: | Jim Nasby <Jim(dot)Nasby(at)BlueTreble(dot)com> |
---|---|
To: | Jake Nielsen <jake(dot)k(dot)nielsen(at)gmail(dot)com>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Cc: | <pgsql-performance(at)postgresql(dot)org> |
Subject: | Re: Unexpected expensive index scan |
Date: | 2016-09-30 22:19:59 |
Message-ID: | 7abde325-3452-a1e1-e288-ee33408d7708@BlueTreble.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On 9/28/16 1:11 PM, Jake Nielsen wrote:
> Beautiful! After changing the random_page_cost to 1.0 the original query
> went from ~3.5s to ~35ms. This is exactly the kind of insight I was
> fishing for in the original post. I'll keep in mind that the query
> planner is very tunable and has these sorts of hardware-related
> trade-offs in the future. I can't thank you enough!
Be careful with setting random_page_cost to exactly 1... that tells the
planner that an index scan has nearly the same cost as a sequential
scan, which is absolutely never the case, even with the database in
memory. 1.1 or maybe even 1.01 is probably a safer bet.
Also note that you can set those parameters within a single session, as
well as within a single transaction. So if you need to force a different
setting for a single query, you could always do
BEGIN;
SET LOCAL random_page_cost = 1;
SELECT ...
COMMIT; (or rollback...)
--
Jim Nasby, Data Architect, Blue Treble Consulting, Austin TX
Experts in Analytics, Data Architecture and PostgreSQL
Data in Trouble? Get it in Treble! http://BlueTreble.com
855-TREBLE2 (855-873-2532) mobile: 512-569-9461
From | Date | Subject | |
---|---|---|---|
Next Message | Jim Nasby | 2016-09-30 22:49:02 | Re: Millions of tables |
Previous Message | Joe Proietti | 2016-09-30 13:44:40 | Re: MYSQL Stats |