Real-time Guarantees for High-Performance Safety-Critical Applications

Abstract

Spacecraft on-board computers (OBCs) are expected to ensure that mixed-criticality workloads meet their timing constraints, as missed deadlines in safety-critical tasks can jeopardize mission safety. A common strategy is to allocate hard real-time and high-performance tasks to separate computing nodes. However, this increases power consumption, cost, and integration effort—resources tightly constrained in space systems. A dual-kernel approach, in which a real-time microkernel runs alongside Linux on the same hardware, offers a way to provide real-time guarantees without dedicated real-time nodes. This work assesses whether Xenomai 4, a dual-kernel Linux extension integrated into the Tasking Framework—a software library for on-board data handling developed at the German Aerospace Center— can enable a single executable binary to run both real-time and high-performance tasks on the same hardware. To this end, a Xenomai-based execution layer was implemented within the Tasking Framework and evaluated on a Xilinx Zynq UltraScale+ MPSoC ZCU104 board. This approach contrasts with traditional split deployments in the Tasking Framework, where Linux handles high-performance tasks and RTEMS handles real-time tasks. The evaluated case study showed reductions of up to 68% in execution time and 86% in jitter compared with a standard Linux setup on the same hardware. These results suggest that real-time guarantees for high-performance applications are achievable on a single platform, potentially reducing hardware duplication and integration effort. These findings motivate further testing and validation.