Dependency Management Framework – a Tool for Designing Robust Real-time Systems

Due to the complexity and requirements of modern real-time systems, multiple teams must often work concurrently and independently to develop the various components of the system. Since a team typically only knows the dependency relations between the components they wrote and those they directly use, keeping track of system-wide dependency relations is not possible for any individual team. To further complicate matters, dependency relations often change as software components are refined or their interactions modified. Because the robustness of any real-time system hinges on the availability of essential services in spite of faults and failures in useful but non-essential components, keeping track of the constantly evolving dependency relations between the system’s components is crucial. If a system’s designers cannot ensure that critical services only USE but do not DEPEND ON less critical components, a seemingly minor fault can propagate along complex and unforeseen dependency chains and bring down the entire system. Therefore, automatically tracking and analyzing system-wide dependency relations given only local dependency information is vital for the development of robust real-time systems. This thesis presents Dependency Management Framework (DMF), a unified theoretical framework and prototype toolkit for dependency management in robust real-time systems. To illustrate the usability and scalability of DMF, we also give several application examples of DMF, with case studies on a distributed car control testbed and a student-developed ION CubeSat satellite.