HA! How many sides of your mouth can you lie out of? You say you like lunix because it's open and Oracle, but now you're going to switch to the mother of all lockins IBM? You're nothing but a total fraud, obviously.
I don't know about that. pSeries are granular to 1/10th of a CPU core, with 1/100th CPU core increments, with an upper limit of 256 microparitions per server. Solaris zones granularity can be share based, where a share is user defineable. More than 100 shares can be defined per CPU. The upper limit for Solaris zone is 8,192 zones per Solaris instance, or 147,456 zones in a single server running 18 instances of Solaris.
This guy is the king of Solaris zones. He has run 600 zones running Apache on a four CPU core server.
Network sharing between zones is handled by Solaris' IP Quality of Service feature. Memory in the Solaris zone construct is shared, but real and virtual memory can be capped per zone.
I don't know much about how ZFS' new feature align with zones, or just how storage allocation and sharing between zones works, however storage management tools like Veritas Volume Manager and Solaris Volume manager, as well as Solaris zones' local filesystem feature should make this pretty easy.