Recovery for sporadic operations on cloud applications

During sporadic operations (e.g. deployment, upgrade, and reconfiguration, etc.) on cloud applications, errors could happen and make the operations fail. This is due to several reasons such as cloud APIs uncertainty. One way to handle the errors is to recover from them. Some existing recovery approaches such as exceptions handling mechanisms usually do recovery by a graceful operation exit or issuing waiting time, and they are faced with the challenge of handling cross-platform and cross-language exceptions. Our research proposes a non-intrusive recovery method to recover from errors during sporadic operations on cloud applications. First, based on analyzing an operation as a process we determine the recovery points according to certain criteria. Second, optimal recovery action will be triggered if errors after any recovery point are detected. The optimal recovery action is selected from eight recovery patterns (such as RewindR sporadic operation; recovery;