Multi-layered Adaptation for the Failure Prevention and Recovery in Cloud Service Brokerage Platforms

Abstract

Self-adaptation is a basic capability of modern applications, which adjust their structure and behaviour at run-time, adapting to changes in their environment, in order to maintain the quality of service at runtime. Models@run-time is an emerging approach for adaptation, whereby a models@run-time engine maintains a causal connection between an application model and the running application, so that a reasoner can adapt the application structure and behaviour by reading and writing this model. However, when used on the dynamic quality control of cloud-based applications, a traditional first-order adaptation is usually not sufficient. This is because during the application life cycle, its requirements may also change, which requires adaptation on the reasoner itself. In this paper, we propose a multi-layered models@run-time approach to enable a second-order adaptation. By maintaining a causal connection between an adaptation model, which reflects the behaviour of the reasoner, and the running application, we enable the adaptation to be automatically adjusted according to the changes in the running application. We apply this approach on a case study for failure prevention and recovery in cloud service brokerage platforms..