Hi!
During recent TechEd sessions there were some talks about Storage Spaces and how to scale out SoFS cluster (for example, http://channel9.msdn.com/Events/TechEd/Europe/2013/MDC-B218#fbid=F3Zyz8qjfcO, 28 — 35 min). It was shown that we can build a cluster using 4 servers and a common JBOD - it makes up 1 unit. Then we can add more such units in the same storage cluster.
The interconnection inside one unit is SAS, the interconnection between units is Ethernet. And it works as one file server cluster, it can survive the failure of the whole unit. That means that data can be mirrored across different units. If so, what prevents us from having one server and one JBOD per unit? And if so, why can't we use local SAS disks instead of JBOD for each server?
The picture at 15 min shows this 1-server-per-unit scenario but all the documentation I could found implies that it's infeasible.