I'd very much like to see what algorithm these systems are using to enable transactions in a distributed environment. Are they just using straight two-phase com

I'd very much like to see what algorithm these systems are using to enable transactions in a distributed environment. Are they just using straight two-phase commit, and letting the whole transaction fail if a single server goes down? Or are are they getting fancy and doing some kind of replication with consensus?