We wrote about the attempt to build a Flashmob supercomputer here , back in February.. The event took place earlier this week, and succeeded in networking nearly 700 computers on the spot; however, they failed at their goal of getting a spot on the list of the Top 500 hundred supercomputers because of (all together now) problems of scale: Dealing with, uh, flakeley nodes is one of the big design challenges of the era, in all sorts of systems. If we're going big and distributed (which we are) then you cannot _ever_ assume things will go your way, certainly not at all points in the system at the same time. Big distributed anything -- supercomputing, file sharing, social networks -- all has the same core challenge: assume flakeliness, then design systems that can withstand it.