Contention Behind the Wheel: Characterizing and Mitigating Interference in Next-Gen Automotive MCUs
Year: 2026
Type: conference-paper
Abstract: The evolution of automotive systems, and in particular the adoption of zonal architectures and controllers, is driving a shift toward platform consolidation, where multiple functions of different criticality levels must execute on a shared hardware platform under tight Size, Weight, Power, and Cost (SWaP-C) constraints. Traditional low-end microcontroller units (MCUs) provide timing predictability but limited computational capability, whereas high-end application processing units (APUs) offer higher performance at the cost of more complex shared microarchitectural hierarchies that can undermine temporal isolation. Across multiple ISAs and vendors, a new class of real-time MCUs is emerging to bridge this gap by combining multicore execution, safety-oriented features, and real-time isolation. However, the interference effects introduced by resource sharing in this new class of platform remain poorly understood. In this paper, we investigate interference in next-generation automotive MCUs through an experimental study on a representative platform, the NXP S32Z270. Our results show that contention in shared on-chip memory resources can induce up to 38% execution-time slowdown and amplify execution-time jitter by up to 153.63×. We further show that memory contention affects interrupt handling, increasing average interrupt latency by up to 1.52× and its jitter by up to 27.2× relative to standalone execution. To mitigate these effects, we propose a lightweight per-core memory bandwidth reservation mechanism that bounds memory pressure from co-runners. The mechanism reduces memory-induced slowdown to below 10%, restores timing predictability, and reduces interrupt-latency overhead and jitter to near 1.06× and 2.72×, respectively.
Authors: Afonso Santos, Diogo Costa, Sandro Pinto
Institutions: Centro ALGORITMI, University of Minho
Language: English
Topic: Real-Time Systems Scheduling
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7204575596