Just looking at the design of RAC, you're still just increasing the n at which it stops scaling. Anything with a bottleneck (in this case, the shared disk image

Just looking at the design of RAC, you're still just increasing the n at which it stops scaling. Anything with a bottleneck (in this case, the shared disk image and the concurrent modification algorithms) is going to wedge for some n.