Smart roadside infrastructure (SRI) and software-defined vehicles (SDVs) are key building blocks of cooperative intelligent transportation systems (C-ITS), where edge systems must meet stringent requirements for latency, throughput, and isolation. Con...
Smart roadside infrastructure (SRI) and software-defined vehicles (SDVs) are key building blocks of cooperative intelligent transportation systems (C-ITS), where edge systems must meet stringent requirements for latency, throughput, and isolation. Containers are widely adopted because they are efficient and easy to deploy, but their comparatively weak isolation raises serious security concerns. Micro virtual machines (microVMs) offer VM-grade isolation with container-like lightweight execution. However, their network performance under cooperative driving workloads remains poorly understood. This paper presents a measurement study of microVMs in edge-assisted cooperative autonomous driving systems. We design a suite of benchmarks based on representative network communication patterns and compare microVMs with containers. The benchmarks cover periodic streaming, request–response transactions, co-located interference, and MQTT-based publish–subscribe messaging. Our results show that microVMs can match container-level performance—and in some cases even exceed it—though the additional virtualization layer increases CPU utilization. We also find that the magnitude of this trade-off depends on both platform design and workload characteristics. Overall, microVMs are feasible for these workloads, but they should be deployed in a workload- and resource-aware manner rather than treated as a direct replacement for containers.