Fair Bandwidth Sharing among Virtual Machines in a Multi-criticality Scope

Abstract

System virtualizations consolidation in separatedvirtual machines provides a reasonable way to integrate formerlydistinct systems into a single mixed-criticality multi-coresystem. We propose an adaptive resource management scheme forvirtualization-based systems that have to be certified. Periodicservers and the elastic task model combine analyzability atdesign time with adaptability at runtime. A mode change or theenabling / disabling of tasks trigger a resource redistribution,which guarantees that a specified minimum is always allocatedand obtains a fair distribution of spare capacity among thevirtual machines. The partitioned scheduling and the assignmentof static priorities ease certification. The scheme has the potentialto improve the resource utilization and support adaptive and selfoptimizingapplications with strongly varying execution times.