If you want to keep voice traffic to a minimum on WAN and if uninterrupted Messaging service to remote locations, that is, local survivability, is important to you, the distributed model might be more appropriate for your organization. The distributed implementation model itself is a form of redundancy. To ensure continued service, you can add an extra application server for redundancy at locations where local survivability is critical.
With the distributed model, if an application server at a remote location cannot reach the central storage server, the caller experience at that location is the same as if the storage server was reachable. Call answering continues to function and users can still access their most recent messages. For more information about how Messaging continues to function when the storage server is unreachable, see Local survivability.
The principal benefits of a distributed topology are local survivability and less voice traffic on the network. The tradeoff is a slightly higher total cost of ownership.