Compositional framework for real -time embedded systems

An embedded system consists of a collection of components that interact with each other and with their environment through sensors and actuators. Two key characteristics of embedded systems are real-time and resource-constrained. As embedded systems become more complex due to increased functionalities, it is desirable to achieve the compositional design and analysis of resource-constrained real-time systems, i.e., the system-level design and analysis on the timing and resource aspects can be achieved by composing independently obtained component-level design and analysis results. In this dissertation, we propose a framework for this problem. In the proposed framework, we develop techniques for the compositional schedulability analysis of real-time systems through real-time component interfaces. We also develop techniques for supporting the compositional design and analysis of resource-constrained real-time systems by determining the resource use of each task within individual components such that a total cost on collective resource use is minimized subject to the system\u27s timing and resource constraints. In the real-time systems community, compositional schedulability analysis has not been adequately addressed except trivial cases. In this dissertation, we extend the results of traditional real-time scheduling theories by including a notion of real-time resource model into schedulability analysis. We propose a periodic resource model that can specify periodic behavior in resource allocations and develop exact schedulability conditions with the worst-case resource supply scenario of the proposed periodic resource model. Based on this result, we derive a periodic component interface that specifies the minimum periodic resource requirements necessary to guarantee the schedulability of individual components. We then achieve the compositional schedulability analysis of real-time systems through the periodic component interface. Typical scarce resources for real-time embedded systems include energy for battery-operated systems and memory for cost-sensitive systems. Many techniques have been proposed to reduce energy consumption and program code size, respectively. These techniques often produce tradeoffs between reducing resource consumption and increasing program execution time. Given such tradeoffs for resource-constrained real-time systems, we consider a multidimensional optimization problem that is to determine the resource use of individual workloads while a total cost on the resource uses, in terms of processor utilization, code size, and processor energy consumption, subject to the system\u27s real-time and resource constraints. Showing the NP-hardness of this problem, we develop a framework for exploring tradeoff space to find sub-optimal solutions efficiently and extend the framework for addressing the problem compositionally. In this dissertation, we propose a framework for supporting component-based design and analysis on timing and resource aspects. Our proposed framework lays a groundwork for future advances of component-based design and analysis techniques for real-time embedded systems