Self-healing framework for Cloud-based services

Cloud computing enables the automated creation of business applications from independently developed and deployed services. In this paper, we propose a technique for Cloud-based services to dynamically evaluate the performance of services based on their previous history, and user requirements. The likelihood of fault occurrence is then used to create recovery plans. The `best' recovery plan is then chosen to be either executed immediately, or saved for a later execution. We assess the overall system dependability using the generated plan and the available invocation information of the included services. Experiment results indicate the applicability of our technique, and show the improved performance in comparison with similar existing approaches.