From Brett Estrade on IRC: the UC Berkeley EECS school has a recent paper out that talks about massively parallel systems and the future needs of those platforms. (That’s where DragonFly is going – parallel – and where it’s been – Berkeley.)
A running description of activity related to DragonFly BSD.
That’s not the same kind of parallelism. They are investigating new programming models for what they expect to be the computing platform of the future (chips with 1000s of cores). This is not very related to cache coherent clustering over wan links nor with scaling on smp systems with a few dozens of CPUs with the standard mutual exclusion and lockless algorithms model.