We're designing a distributed fault-tolerant virtual machine, and we're trying to determine the most cost-effective infrastructure setup. Google's infrastructure is famously built from lots of cheap computers that break down all the time (and is supposed to be very efficient from a cost per query perspective), and while there are specs floating around online for their hardware from several years ago, there's a dearth of more recent information.
Does any know where to find the typical specs (disks, memory, processors, etc.) for a new commodity box at Google or another place with a similar distributed infrastructure setup?