ScholarWorks@성균관대학교

초록

As automotive architectures transition to Software Defined Vehicle (SDV) with High-Performance Computer (HPC), Service-Oriented Vehicle Diagnostics (SOVD) emerges to replace traditional Unified Diagnostic Services (UDS). This study compares Native SOVD and SOVD-to-UDS diagnostic paths in a virtualized in-vehicle gateway using the NXP S32G399A platform with Xen hypervisor. Three guest domains were configured representing a gateway, ADAS controller, and legacy ECU simulator. Load testing with 10-100 concurrent connections revealed that SOVD-to-UDS conversion exhibits higher response times than Native SOVD. CPU utilization analysis revealed that Native SOVD maintained substantially higher usage compared to SOVD-to-UDS, indicating I/O wait times as the primary bottleneck rather than computational overhead. These findings quantify the performance cost of legacy compatibility and provide insights for designing optimal hybrid diagnostic architectures in next-generation vehicles.

키워드