Exploring Xen/KVM in prototyping an automotive use-case
Ref: CISTER-TR-191212 Publication Date: 30, Sep, 2019
Exploring Xen/KVM in prototyping an automotive use-caseRef: CISTER-TR-191212 Publication Date: 30, Sep, 2019
Due to increasing autonomy in vehicles, the automotive industry is searching for solutions which allow the reduction of engineering costs resulting from increasing hardware and software requirements. To solve the problem, one of the solutions being studied is the utilization of virtualization to allow multiple systems to coexist on the same hardware, allowing the systems to function as if they are isolated, while reducing hardware costs. Using virtualization, the system becomes capable of hosting services of mixed criticality safely, reducing interference between the services through temporal and spatial isolation. In this project we analysed the Xen hypervisor in an effort to better comprehend how its inclusion would impact the execution of a system with real-time requirements. After a brief look at the current state of the art we describe how Xen behaves on the 32-bit ARM architecture, how it shares peripheral devices between virtual machines and how interdomain communication can be achieved. Following that, we present an example of how to deploy Xen on a Banana Pi SOC. After studying how best to establish a continuous connection between different virtual machines we tested our solution for message transfer jitter, deadline failure rate impact and deployed a prototype of an automotive application to verify how the system would behave in a real-world scenario. As planned, a system of mixed criticality was deployed on the hardware using the Xen hypervisor, PREEMPT RT was deployed on one virtual machine with real-time requirements, while Linux was deployed on a general-purpose virtual machine. The test showed that on the selected hardware Xen was unable to guarantee temporal isolation, showing significant performance drops in deadline failures. The communication between virtual machines and the deployment of the real-world application, however, were successful.
BEng Thesis, ISEP.
Notes: Orientação científica: David Pereira, Coorientação: Cláudio Maia