Towards Modeling and Validating Analysis Processes for Software Adaptation

With the complexity and the dynamic analysis requirements during software adaptation for software in the open and dynamic environment, the software analysis requires integrating multiple analysis methods and controlling the execution of analysis methods in specific sequence. As a result, the analysis process is hard to be communicated to its designers, implementers and users without an explicit model. In addition, if dependent methods are not included or encounter runtime failure, the data dependency and control dependency between analysis methods may cause correctness problem. In this paper, we proposed a domain-specific language for modeling the analysis process using subset of notations from BPMN. In order to cope with the dependency problem in model and implementation level, the dependency relationships are divided into static dependency and dynamic dependency. Dependency identification and matching algorithm is provided to assure the static dependency in the analysis process model. For the dynamic dependency, we provide a service to detect the analysis failure caused by dynamic dependency. We model and validate two analysis processes supporting software adaptation with performance and reliability concerns, and apply the analysis processes in analyzing a running system ECPerf. Keywordssoftware analysis; analysis process, process validation, analysis process modeling