A Pattern Language for Manual Analysis of Runtime Events Using Design Models

Abstract

Modeling is an important activity in the software development process whose output are design artefacts that describe the resulting software from a high-level perspective. Recent research investigates the role of models at runtime and the results indicate that analysts perform better at observing the behaviour of a running system if they can utilize models during the analysis. However, setting up a system which allows the analysis of its behaviour at runtime using models involves many challenges regarding the modeling environment, the introspection infrastructure, the traceability management and the analysis integration. This paper summarizes design alternatives for implementing systems with manual analysis support by investigating recurring concepts like patterns, modeling habits, languages, middlewares and development techniques found in approaches that utilize models at runtime. We organize the gained knowledge as patterns in a pattern language which captures various issues and their solution alternatives, including their benefits and liabilities. The pattern language consists of modeling patterns for setting up the models and the environment for the analyst, introspection patterns for extracting data from the running system, traceability patterns for relating the extracted data with the models, and analysis patterns for processing the extracted data using the models. We demonstrate the application of the pattern language based on the implementation of a robot system.