Every part of the system is redundant and distributed over two data centers per region. There is no single point of failure. Regard the redundancy logic of your access network.