This topic describes the checkpointing process, including:
The backup application is responsible for checkpointing circuit state information to the ISUP implementation running on the backup board. This information consists of the call processing state and circuit blocking state. This is the minimum information to be checkpointed from the primary application to backup application.
Upon receipt of checkpoint information, the backup application must checkpoint the circuit state information (call processing and circuit blocking states) to the ISUP stack on the backup board. This is accomplished through a call to ISUPStatusReq with an eventType of CIRGRPSET.
Optionally, the primary application can checkpoint the transient state of a circuit. A circuit is in a transient state during call setup and call release. Checkpointing the transient state of a circuit allows the backup application to identify the circuits that were in call setup or release at the time of changeover. This information is not checkpointed to the ISUP stack.
Note: After changeover, the backup (now primary) application resets all circuits in a transient state.
Incremental checkpointing refers to the transmission of state information for a single circuit concurrent with a change in state for the referenced circuit.
This is the recommended checkpoint method for the following reasons:
Backup application always has up to date information concerning all circuits.
Transient circuit checkpointing is reasonable due to the timely update of information.
Minimum number of circuit resets upon changeover.
Acceptable for all size configurations.
Batch checkpointing refers to the mass transmission, from primary application to backup application, of state information for controlled circuits and hardware.
Periodic batch checkpointing refers to the periodic transmission of state information for all controlled circuits and hardware.
Periodic delta batch checkpointing refers to the periodic transmission of state information for controlled circuits and hardware that have changed state since the previous checkpoint. This is more efficient than normal periodic batch checkpointing.
The use of batch checkpointing is discouraged for several reasons:
Large window of uncertainty; circuit activity between checkpoints is not visible to backup.
Depending on the time between checkpoints, transient circuit checkpointing may not be effective.
After changeover, transient circuits are unusable until protocol recovery or until all idle circuits are reset by the new primary.