On Wed, Jan 28, 2009 at 10:39:51PM -0800, Jason Bradfield wrote:
> I found the what was causing the difference...
> 
> Most of the user time was spent in 'my_utf8_uni' which from what I
> gather a routine to convert utf8 to latin1(or unicode)
> 
> Once I converted all charsets on our database to latin1 our
> performance became the same...  This lib call on ubuntu must be more
> efficient than on soalris..

Are you running in a UTF-8 locale on Ubuntu?  I bet you are on Solaris.

Is this function being called both, on Ubuntu and Solaris?

> Also, buy converting the database to latin1 we saw this one sql drop
> from 19 seconds to 13 seconds, this is quite a significant
> difference...

UTF-8 for *collation* slows everything down.  ls(1) in a directory with
a million files will go reasonably fast in a C locale, but will take
forever in a UTF-8 locale.

Nico
-- 

Reply via email to