Hierarchical Scheduling for Plug-and-Produce

Today's increasing number of sensors and computation nodes covered by cyber physical systems (CPS) results in rising complexity. For instance, computation results are based on measured data whose quality strongly depends on their age. CPS therefore have real-time requirements on computation results and communication to keep temporal dependency between measured inputs and computed outputs. In addition, today's CPS shall be modular to enable flexibility and scalability, e.g., postulated for production systems in context of Industry 4.0. Enabling CPS to easily integrate components by some Plug-and-Produce mechanism is desired.In this paper, we aim at enabling Plug-and-Produce in CPS using hypervisor-based virtualization. This implies hierarchical scheduling of dependent real-time systems. Here, dependencies are given by precedence constraints of tasks. Based on an approach for detection of new components added to a real-time network, in this paper we focus on integration of enabled applications into the current schedule of a computation node. Here, enabled application refers to an application software that just got executable by plugging some component to the CPS. Applications are encapsulated by virtual machines and provide a self-description including information about required and provided data as well as timing behavior. This self-description is used to adjust global scheduling and thus include new functionality to the CPS.