Redundency for high-availaibility. If a node fail, I can restart the service manually, or automaticly on another node, without loosing any data. Also, there are come common data between services that need to be availaible in real-time. > If that's the case, why look at shared storage at all?