at vBSDCon 2013 they explained the immense redundancy and diversity of their NS hosting of .com, .net, and .org. At their POPs they have: * Diverse switches

at vBSDCon 2013 they explained the immense redundancy and diversity of their NS hosting of .com, .net, and .org. At their POPs they have: * Diverse switches * Diverse load balancers * Diverse servers (from different mfgrs AND hardware generations) * Diverse OSes (Linux & FreeBSD) * Diverse DNS software: BIND, NSD, and ATLAS I don't know how they decide the exact configurations here, but the idea is that a single vulnerability or bug is unlikely to cause a full outage at the POP. The complexity here is immense and they must have a really great system for managing the configurations and replicating DNS changes. It certainly would not be cheap to build.