A Pattern-Oriented Design Framework for Self-Adaptive Software Systems

Abstract

Multiple interacting MAPE-K loops, structured according to specific interaction patterns, have been introduced to design the adaptation logic in case of decentralized self-adaptive software systems. Designing such complex systems requires the availability of tools where MAPE patterns can be easily instantiated to provide fast architectural solutions, and the encoding towards specific domains is facilitated by automatic mapping of such pattern instantiations in domain-specific languages; validation and verification must be also supported to assure correct development of reliable systems. In this paper, we present a pattern-oriented framework, based on the MSL (MAPE Specification Language) modeling language, for the design of self-adaptive systems. The framework supports: (i) explicit modeling of the adaptation logic in terms of patterns of interactive MAPE-K loops; (ii) ability to tailor MSL models for a specific application domain and synthesize from them other modeling artifacts/code according to a target implementation context and scope (e.g., OpenHAB); (iii) ability to perform validation and verification of MSL models by means of the ASMETA formal framework.