by mhonman » Mon Sep 23, 2013 5:54 pm
Yep, Parallella is the polar opposite of a virtualisable system. In the latter case the problem is lots of server instances that are mostly idle (legacy of Windows NT server deployment practices - to keep it stable each function had to run on its own box). So buy one big box and consolidate them onto it as VMs. Saves loads of server room space and power.
In the case of Parallella it is a tiny device meant to do a lot of calculations at the same time, and very cheaply too. No space/power or "server sprawl" problems, hence no value in virtualisation support. In the case of Parallella the more interesting question is how many of them can on hook together to solve a single problem, rather than how many ways can one divide a single Parallella.
If you do want to play with virtualisation, the best way is with a secondhand server that supports KVM (noisy, though!).