Re: [PERFORM] performance contradiction - Mailing list pgsql-performance

From Feike Steenbergen
Subject Re: [PERFORM] performance contradiction
Date
Msg-id CAK_s-G3sARA1-450K91HfHtMtZJRCzHPcGKtR9d_vDuAWyA8ag@mail.gmail.com
Whole thread Raw
In response to [PERFORM] performance contradiction  (Gabriel Dodan <gabriel.dodan@gmail.com>)
List pgsql-performance
On 23 January 2017 at 17:55, Gabriel Dodan <gabriel.dodan@gmail.com> wrote:
>
> BUT if I run a trivial select on both servers, on a similar table, the select
> perform much much better on second server!

You're comparing two very different systems it seems, therefore you might be
looking at difference in the performance of EXPLAIN, just getting timing
information of your system may be the most expensive part[1], you could disable
the timing explicity:

EXPLAIN (ANALYZE ON, TIMING OFF) <query>

And, there is something that stands out:

So it seems there is also some difference in the data, we could validate the
actual numbers:

SELECT sum(pg_column_size(symbols_tests))/count(*) FROM symbols_tests;

regards,

Feike

pgsql-performance by date:

Previous
From: PC Drew
Date:
Subject: [PERFORM] Inaccurate GiST Index Cost Causes DB Contention
Next
From: Mike Beaton
Date:
Subject: [PERFORM] Correct use of cursors for very large result sets in Postgres