Towards Hierarchical Scheduling of Dependent Systems with Hypervisor-based Virtualization

Multicore hardware platforms raise more and more interest in cyber physical systems due to the increasing computational power demands. In addition, multicore systems promise reduction of network complexity by enabling composition of former singlecore entities onto a single network node. Hypervisor-based virtualization is well known to enable running various software stacks concurrently on a single hardware platform as well as porting singlecore software to a multicore system. Depending on the kind of applied virtualization - full or para-virtualization - porting software requires no or just low adaptation effort. However, software within cyber physical systems is characterized by the need of communication that has to be served even though it is running in a set of virtual machines. Typical hardware platforms addressed by virtualization approaches are multicore systems enabling not only concurrent but even parallel execution of guest systems. In this paper, we propose a concept to compute a feasible global schedule for dependent systems executed in various virtual machines on a multicore hardware unit. To enable transfer of legacy software from single-to multicore platforms, we apply hierarchical scheduling and thus enable keeping scheduling policy of original software untouched. As a first step towards hierarchical scheduling of dependent systems, we consider guest systems scheduled by EDF* to show capabilities of our concept. Future work aims at extending our concept for further local schedulers.