The redundant application models are:
Dual-node (distributed)
In a single-node application model, both boards reside in the same chassis and a single application instance opens and simultaneously binds to the same SCCP user SAPs on both boards.
At any given point in time one board is primary, the other is backup. The single-node application tracks which board is primary and directs all traffic toward the primary board.
Because a single application instance handles all traffic to and from both boards, no external mechanism is required to synchronize state information between application instances.
In a dual-node application two instances of the same application exist, typically on separate chassis. Each instance of the application binds to a single board and, at any given point in time, one instance of the application is the primary and one is the backup (corresponding to the board states).
Only the primary application instance sends and receives network traffic; the backup monitors the state of the primary, waiting to become primary itself.
In this model, some external mechanism may be needed to synchronize the state of the primary and backup application instances - this is up to the application. For example, if connection-oriented services are employed, the primary application instance may need to checkpoint the connection state to the backup application instance so the backup can preserve active (confirmed) connections across a switchover.
Note: It is possible to implement a single-node system employing two application instances, one for each board. The considerations for this case are the same as for a dual-node architecture (although the implementation may be different).
In either application model, the application must register for the HMI service and the SCCP service to fully support redundant operation.