Error recovery and retrieval

Error compensation can be achieved by establishing a mechanism for recovery  

Retrieval (back recovery of errors) consists of restoring the system to a previous secure state. To do this, regular backups of the system must be undertaken and one of the backups must be able to be charged. Upon detection of an erroneous situation, it is possible to reload a previous situation and restart the execution. If the error comes from the environment or from a transient failitre, the system should resume proper operation. If this is a systematic failitre (hardware or software), the system will return to an erroneous state. Some systems also dispose of several software application alternatives and can activate another replica of the application during retrieval. 

The main advantage lies in the fact that the erroneous state is deleted and that this deletion is not based on finding the location of the fault or the cause. The back recovery of errors can be used to recover unanticipated faults, including design errors. 

Retrieval can be either of high or low grain. The recovery by high grain restores the last overall correct condition and rerans the same software application completely. Recovery by low grain is a finer recovery strategy with the restoration of a local correct state (confined to a function, for example), alternative execution, etc. 

High-grained recovery has the advantage of setting up a single point of recovery and remains suitable for systems subjected only to hardware failures. 

---