On Sat, Jan 29, 2005 at 09:33:44PM +0100, Marcin wrote:
> 
> The second option is Slony-I. I gave it a try, and it seems to work 
> nicely. However, before I go into deep tests, I'd like to know if I 
> could get a better performance with it than with dbmirror (as it's also 
> trigger based)? Does anyone use Slony-I in similar environment? Perhaps 
> in my setup it's better to go with PITR?

I don't know, because we never tested dbmirror.  What I do know is
that Slony in our tests was at least a factor of 3 faster than
erserver.  There _is_ an overhead to Slony, but it's an overhead that
as far as I can tell is under 10%.  

I should note that the Xeon issues may be part of what's biting you
here.  The more update transactions you have to do (== the more locks
you have to acquire) the more likely the Xeon spinlock issues will
bite you.  In your shoes I would look seriously at replacing the
hardware.  It might be your simplest answer.

A

-- 
Andrew Sullivan  | ajs at crankycanuck.ca
A certain description of men are for getting out of debt, yet are
against all taxes for raising money to pay it off.
		--Alexander Hamilton