Solaris can manage this on Sun's 1000+-level Ultra Enterprise servers.
(I think... It's available in the StarFire (UE10000), and I'm fairly
certain the capability has just been added to the others.) It's quite
handy in principle; you can do server failover within a single box. I
have no idea if UltraPenguin handles this... I would be floored if it
did, as it's a fairly recent addition to main-line Solaris.
I'm not sure if the PC bus-handling chipsets could do this well, though.
The I/O APIC's redirection of interrupts might make it possible (slowly
reading up on it), but I don't know enough yet.
- The interesting case that occurred to me would be if another OS [...]
Someone will have to arbitrate access to whatever is used to point
interrupts to different processor groups. Your suggestion of running
an MP-unaware OS is interesting, but I'd be much more interested
(personally) in multiple copies of Linux with failover support. That
would need a _lot_ of help from the bus controllers.
The complexity involved in making it work well hurts my head. And I
don't know enough about the various SMP chipsets even really to be
saying much, so I'll shut up now.
Jason
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu