Performance Analysis of KVM Hypervisor Using a Self-Driving Developer Kit
Thilo Muller, Hadi Askaripoor, Alois Knoll · IECON 2022 – 48th Annual Conference of the IEEE Industrial Electronics Society · 2022
Virtualization plays an increasingly important role in safety-critical embedded systems like modern vehicles due to its numerous advantages like cost reduction, better scalability, or easier safety verification. These advantages come at the cost of additional performance overhead. While performance evaluation of virtualization solutions for server applications has been the subject of research for a long time, it is still not adequately investigated for embedded use cases. We present the first work to analyze the performance of KVM, an open-source hypervisor, on an Nvidia Drive AGX, a developer kit for self-driving vehicles. Experimental measurements show that the computational overhead restricts the number of partitions. Also, poor memory and file I/O performance suggest that KVM is not usable for practical applications with the Nvidia Drive AGX.