Comparative evaluation of offset scheduling methods for AFDX networks

2026-9-3
Çelik, Kaan
AFDX (Avionics Full-Duplex Switched Ethernet) is a flow-based, fully switched avionics network standard. Its flows, called Virtual Links (VLs), are traffic-regulated to support time-critical communication. Network Calculus computes worst-case end-to-end delay bounds. Such bounds may become optimistic when legitimate VL configurations violate analysis assumptions, or highly pessimistic when known temporal relations among transmissions are not exploited. This thesis revisits an existing offset-aware Network Calculus analysis for periodic VLs with source offsets, corrects its treatment of source-timing information under general routing, and refines it to obtain tighter VL-specific bounds. Our first contribution shows that same-source VLs that diverge and later reconverge can invalidate source-derived serialization relations and lead to optimistic bounds. We introduce route-coherent source groups, which retain these relations only when justified by the VLs' common route history. The second contribution reduces the remaining aggregation pessimism through VL-specific analysis. For the source group containing the target VL, we retain the exact frame sequence at the first switch and construct conservative compacted sequences downstream, while other traffic remains bounded by safe arrival curves. We establish the correctness of both formulations analytically and evaluate them using controlled examples and large realistic configurations. The experiments confirm that the missing route-coherence condition in the existing method can lead to optimistic bounds. The VL-specific analysis produces tighter bounds for up to 99.9% of paths, with mean path-level reductions of up to 31.92%. Overall, route coherence is necessary for correctness, while preserving target-specific timing information significantly reduces pessimism in large AFDX networks.
Citation Formats
K. Çelik, “Comparative evaluation of offset scheduling methods for AFDX networks,” M.S. - Master of Science, Middle East Technical University, 2026.