Detailseite
Projekt Druckansicht

Architektur und Mechanismen des Multi-Change Control Layer (MCCL)

Fachliche Zuordnung Rechnerarchitektur, eingebettete und massiv parallele Systeme
Förderung Förderung von 2013 bis 2020
Projektkennung Deutsche Forschungsgemeinschaft (DFG) - Projektnummer 206480214
 
Erstellungsjahr 2020

Zusammenfassung der Projektergebnisse

The A1 project was engaged in developing the basic (software) architecture and the mechanisms for CCC to control changes of application and platform components. In this central role, A1 provided a networked run-time environment (RTE) for the C projects. This RTE was based on the Genode OS Framework that strictly applies the concepts of component-based operating systems and serviceoriented interfaces. Furthermore, A1 developed the MCC core models and algorithm for applying the methods developed in the B projects. In this regard, an overarching and modular framework for automated model-based integration was developed. There were three major scientific results that not only contributed to the MCC automation concept but are also relevant far beyond the application of CCC. A first accurate timing model was developed in order to enable worst-case response-time analysis for service-based architectures. Those architectures are receiving an increasing interest for shifting from statically configured systems to a dynamic/adaptive system configuration in which software components provide services to which other components can connect at runtime. Prominent examples for this are the robot operating system (ROS2) and Adaptive AUTOSAR. In conjunction with microkernels, a strong separation is possible on component-level, which is required for mixed-criticality systems. From a timing perspective, these systems substantially differ from the classical periodically executed task set used by the real-time research community. A major result was the task-chain model and analysis that accurately reflects the service-oriented communication by incorporating the particular precedence and blocking effects. An automated software configuration environment that is controlled by constraints on multiple design layers that builds the core MCC framework. This framework incrementally refines a platformand implementation-independent specification to a specific system configuration by automating design parameter decisions. Constraints are checked by admission tests. A backtracking algorithm allows an iterative design-space exploration if an admission test fails by revising the responsible decisions. Although parameter decisions and admission tests are assisted by constraint solving methods, we found that these methods cannot be efficiently applied for solving the configuration problem holistically. A self-aware monitoring environment to enforce multi-viewpoint contracts. A first technical contribution for closing the model-implementation gap by platform-centric self-awareness was done within A1 w.r.t. scheduling. We provide temporal isolation and enforcement by budget-based scheduling in the presence of uncertainty of required budgets (from software components) and scheduling overhead (from OS/kernel). By monitoring scheduling overheads and execution budgets, a long-term adaptation of models to the actual (observed) platform behaviour was enabled. CCC work on self-awareness influenced research far beyond the research group. Several related papers and special sessions in top journals and events, organized and authored with external partners, were strongly influenced by the results of A1. One major collaborative result of A1 was the demonstrations of the D1 and D2 showcases at the “Autonomous System Design” workshop and exhibition at the DATE19 conference in Florence, Italy. In both cases, A1 provided the software platform.

Projektbezogene Publikationen (Auswahl)

 
 

Zusatzinformationen

Textvergrößerung und Kontrastanpassung