>
>
>
>	AFAIK, that involves being able to run on a ccNUMA architecture.
>

There's a difference between running on NUMA, and running *efficiently* 
on NUMA...