Automatic Deployment of Embedded Real-time Software Systems to Hypervisor-managed Platforms

Abstract

The deterministic integration of concurrent func- tions on shared multicore platforms is a challenging yet important task. Especially in safety-critical environments, hypervisors can be used to achieve time and space partitioning, but their sole application is often insufficient to guarantee deterministic timing and data flow behavior. Considering the growing complexity of modern embedded systems, for example in terms of functionality and mixed-criticality requirements, model-based approaches are a promising starting point to tackle this issue. In this work, we bridge the gap between a model-based behavior specification methodology based on the Logical Execution Time (LET) concept and target platforms running a commercially available bare- metal hypervisor. Therefore, this paper describes a runtime environment that implements LET semantics at the level of hypervisor partitions and a tool-supported design methodology that deploys software to this runtime environment. From a be- havior specification provided as a system model with annotated C code, the presented deployment tool generates binary images with guaranteed timing and data-flow behavior for the XtratuM hypervisor. The approach is finally validated by applying it to a Flight Assistance System (FAS) from the avionics domain.