Self-Integration for Virtualization of Embedded Many-Core Systems

Abstract

Virtualization of embedded systems enables the integration of various independent and isolated virtual environments on the same hardware. It is therefore seen as a key enabler for implementing mixed-criticality systems on future many-core technology. However, a challenge is that such systems are increasingly exposed to dynamic workloads, e.g., due to the "openness" of modern mobile devices or the execution of varying application mixes. As a remedy, run-time management enables the dynamic allocation of system resources to the virtual environments according to their current requirements. This paper investigates self-integration of such dynamic environments in virtualized embedded systems by providing management strategies for re-partitioning shared resources at run time. We formulate and analyze the underlying optimization problem with the particular focus on security and scalability as two important factors for managing embedded systems in the many-core era. The evaluation of the theoretical results by means of a case study shows the improvements with respect to scalability and security, and leads to a discussion of the most important future trends for paving the road to self-integrating virtualized embedded systems.