Virtualization in multicore real-time embedded systems for improvement of interrupt latency

Abstract

This paper investigates the possibilities of using virtualization in multicore real-time embedded systems based on combination of general-purpose and real-time operating systems running on separate processor cores. The research is focused on ARM processor architecture due to its widespread usage in numerous application domains. The paper describes the concepts and practical considerations of using virtualization to achieve better interrupt latency both in Linux, as a typical representative of commonly used general-purpose operating system in applications requiring real-time response, and real-time operating system running on separate cores. In the proposed approach we demonstrate how to use virtualization mechanism to decouple cores responsible for real-time and non real-time tasks in a multiprocessor system with real-time requirements. Xen Hypervisor is used as a virtual machine monitor with default credit and new experimental null scheduler. The aim of research was to study the system response time characteristics and to assess practical usability of such approach in applications with hard real-time requirements.