The often huge configuration spaces of modern software systems render the detection, prediction, and explanation of defects and inadvertent behaviors challenging tasks. Besides configurability, a further source of complexity is the integration of cyber-physical systems (CPSs). Behaviors in CPSs depend on quantitative aspects such as throughput, energy consumption, and probability of failure, which all play a central role in new technologies like 5G networks, tactile internet, autonomous driving, and the internet of things. The manifold environmental influences and human interactions within CPSs might also trigger reconfigurations, e.g., to ensure quality of service through adaptivity or fulfill user’s wishes by adjusting program settings and performing software updates. Such reconfigurations add yet another source of complexity to the quest of modeling and analyzing modern software systems.
The main contribution of this thesis is a formal compositional modeling and analysis framework for systems that involve configurability, adaptivity through reconfiguration, and quantitative aspects. Existing modeling approaches for configurable systems are commonly divided into annotative and compositional approaches, both having complementary strengths and weaknesses. It has been a well-known open problem in the configurable systems community whether there is a hybrid approach that combines the strengths of both specification approaches. We provide a formal solution to this problem, prove its correctness, and show practical applicability to actual configurable systems by introducing a formal analysis framework and its implementation. While existing family-based analysis approaches for configurable systems mainly focused on software systems, we show effectiveness of such approaches also in the hardware domain. To explicate the impact of configuration options onto analysis results, we introduce the notion of feature causality that is inspired by the seminal counterfactual definition of causality by Halpern and Pearl. By means of several experimental studies, including a velocity controller of an aircraft system that required new techniques already for its analysis, we show how our notion of causality facilitates to identify root causes, to estimate the effects of features, and to detect feature interactions.:1 Introduction
2 Foundations
3 Probabilistic Configurable Systems
4 Analysis and Synthesis in Reconfigurable Systems
5 Experimental Studies
6 Causality in Configurable Systems
7 Conclusion
| Identifer | oai:union.ndltd.org:DRESDEN/oai:qucosa:de:qucosa:78543 |
| Date | 21 March 2022 |
| Creators | Dubslaff, Clemens |
| Contributors | Baier, Christel, Margaria, Tiziana, Technische Universität Dresden |
| Source Sets | Hochschulschriftenserver (HSSS) der SLUB Dresden |
| Language | English |
| Detected Language | English |
| Type | info:eu-repo/semantics/publishedVersion, doc-type:doctoralThesis, info:eu-repo/semantics/doctoralThesis, doc-type:Text |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | info:eu-repo/grantAgreement/Deutsche Forschungsgemeinschaft/Exzellenzstrategie des Bundes und der Länder/390696704//Cluster of Excellence EXC 2050/1/CeTI, info:eu-repo/grantAgreement/Deutsche Forschungsgemeinschaft/Sonderforschungsbereiche/389792660//Collaborative Research Center TRR 248/CPEC |
Page generated in 0.0067 seconds