Re: Size and performance hit from using UTF8 vs. ASCII? - Mailing list pgsql-performance

From Jeffrey W. Baker
Subject Re: Size and performance hit from using UTF8 vs. ASCII?
Date
Msg-id 1139414042.8707.7.camel@localhost.localdomain
Whole thread Raw
In response to Size and performance hit from using UTF8 vs. ASCII?  (Ron <rjpeace@earthlink.net>)
List pgsql-performance
On Wed, 2006-02-08 at 09:11 -0500, Ron wrote:
> I'm specifically interested in the default C Locale; but if there's a
> difference in the answer for other locales, I'd like to hear about
> that as well.

The size hit will be effectively zero if your data is mainly of the
ASCII variety, since ASCII printable characters to UTF-8 is an identity
transform.  However anything involving string comparisons, including
equality, similarity (LIKE, regular expressions), or any other kind of
comparison (ORDER BY, GROUP BY) will be slower.  In my experience the
performance hit varies from zero to 100% in CPU time.  UTF-8 is never
faster that ASCII, as far as I know.

However, if you need UTF-8 then you need it, and there's no point in
worrying about the performance hit.

You may as well just do two benchmark runs with your database
initialized in either character set to see for yourself.

-jwb



pgsql-performance by date:

Previous
From: Stephan Szabo
Date:
Subject: Re: optimizing away join when querying view
Next
From: Markus Schaber
Date:
Subject: Re: Default autovacuum settings too conservative