Postgres Pro
Downloads
Home   >   mailing lists

Thread: Slow counting on v9.3

Slow counting on v9.3

From
Kai Sellgren
Date:
Hi,

I'm experiecing slow count performance:

SELECT COUNT(*) AS "count"
FROM "NewsArticle"

Takes 210 ms. I've run analyze and vacuum. I'm on 9.3. Here're the stats http://d.pr/i/6YoB

I don't understand why is it that slow. It returns 1 integer, and counts without filters.

This performs quickly:

SELECT reltuples AS count
FROM pg_class
WHERE relname = 'NewsArticle';

But I'd like to add conditions so I don't like the last method.


--
Yours sincerely,
Kai Sellgren

Re: Slow counting on v9.3

From
Mehmet Çakoğlu
Date:
Hi Kai, 

You are right, postgresql Count() function is slow, because; It's physically count the rows one by one. 

Other database systems using indexes for counting, but postgresql walk through all rows in multiple transactions with different row states for calculating the real row count. This is about architecture of postgresql. 

If you use WHERE condition on indexed column in your query, this will be much faster. 







On Mon, Jan 13, 2014 at 11:57 PM, Kai Sellgren <kaisellgren@gmail.com> wrote:
Hi,

I'm experiecing slow count performance:

SELECT COUNT(*) AS "count"
FROM "NewsArticle"

Takes 210 ms. I've run analyze and vacuum. I'm on 9.3. Here're the stats http://d.pr/i/6YoB

I don't understand why is it that slow. It returns 1 integer, and counts without filters.

This performs quickly:

SELECT reltuples AS count
FROM pg_class
WHERE relname = 'NewsArticle';

But I'd like to add conditions so I don't like the last method.


--
Yours sincerely,
Kai Sellgren

Re: Slow counting on v9.3

From
Guillaume Cottenceau
Date:
Kai Sellgren <kaisellgren 'at' gmail.com> writes:

> Hi,
>
> I'm experiecing slow count performance:
>
> SELECT COUNT(*) AS "count"
> FROM "NewsArticle"
>
> Takes 210 ms. I've run analyze and vacuum. I'm on 9.3. Here're the stats http:/
> /d.pr/i/6YoB
>
> I don't understand why is it that slow. It returns 1 integer, and counts
> without filters.

You might actually have a lot more dead tuples than reported in
statistic. Last vacuum is old according to your screenshot. Try
"VACUUM FULL ANALYZE" on your table, then try again counting.


> This performs quickly:
>
> SELECT reltuples AS count
> FROM pg_class
> WHERE relname = 'NewsArticle';

This is not the same. This one uses precomputed statistics, and
doesn't scan the actual table data.


> But I'd like to add conditions so I don't like the last method.
>
>
> --
> Yours sincerely,
> Kai Sellgren

--
Guillaume Cottenceau