One of our core product goals is to actually allow relatively hot swapping of individual compute sleds. It's true that each sled is an x86 server, but there's

One of our core product goals is to actually allow relatively hot swapping of individual compute sleds. It's true that each sled is an x86 server, but there's control software managing the whole rack as a unit, with live migration of workloads and replicated storage. The architecture of the whole thing leans towards treating sleds as essentially like CPU and memory (and storage) boards in previous generations of large scale computers like mainframes.