Georedundancy is an optional configuration provided for non-CMP clusters in which the spare server can be located in a separate geographical location, as shown in Figure 1. The active server replicates state data to the standby and spare servers. If the two servers at one site become unavailable, the third server, located at the other site, automatically becomes the active server and continues to provide service. You can designate sites as primary and secondary.
Georedundancy supports both session-stateful (at the MPE device level) and binding-stateful (at the MRA device level) failover between a pair of geographically separate (or geo-diverse) Policy Management sites. This includes the ability to maintain ongoing sessions and existing bindings that were in progress on the failed site at the time of failure, as well as being able to initiate and handle all new sessions and bindings on the secondary site for the duration of the failure.
- Gateways, content filters, application servers, and other clients are connected to active MRA devices. Each client has a primary connection to the active MRA device at one site and a secondary connection to the active MRA device at the other site. (This is no different than the client connections in a non-georedundant topology.)
- Active MPE devices establish Sh connections, either directly or through Diameter Routing Agents (DRA), to SPRs. The active MPE device at the primary site establishes an Sh connection with a primary IP address, and the spare MPE device at the secondary site establishes an Sh connection with a secondary IP address for use if the spare is promoted to an active role.
- The active MPE devices establish Sy connections, either directly or through DRAs, to online charging servers (OCSs). The active MPE device at the primary site establishes an Sy connection with a primary IP address, and the spare MPE device at the secondary site establishes an Sy connection with a secondary IP address for use if the spare is promoted to an active role.
Using this configuration, if one site fails, clients retain connectivity to the other site, and established sessions remain active. As servers at the failed site recover, they become standby servers, and current state data for the clusters are replicated to them. After the recovered servers are synchronized with the state data of the active servers, they are automatically returned to active roles. This behavior is called revertive which means that if an active server fails and then recovers, it becomes the active server again.
