Extending Mixed Criticality Scheduling

The capability of hardware is constantly developing in capacity, speed and efficiency. This development has sparked industrial and academic interest in how best to utilise the increased capability. It is now possible to integrate many systems that in the past might have existed as different nodes, into the one consolidated architecture. This desire to centralise functionality leads to the potential of a system that contains software components of differing levels of importance or criticality. Such Mixed Criticality Systems pose a challenging problem with regard to analysis and certification. Much work has been undertaken investigating the use of Fixed Priority scheduling for Mixed Criticality Systems, a notable scheme, known as Adaptive Mixed Criticality (AMC), provides significant advances in schedulability over prior approaches. The focus of the work on AMC revolves around just two levels of criticality. In this work we develop extensions to consider greater than two levels of criticality, for both forms of AMC analysis (AMCrtb & AMCmax) and consider the implication of applying these extended approaches. Alongside this we adapt some of the schemes developed prior to AMC in order to assess their relative effectiveness. We also review and further develop Period Transformation for use with Mixed Criticality Systems. Finally we provide a set of evaluations to illustrate the results. We conclude that AMC maintains its effectiveness over many criticality levels and remains an effective scheme. Of the two forms of analysis, AMCrtb is the most practical as the schedulability improvement gained by using AMCmax is slight and the increase in computation required is extreme. When considering an arbitrary number of criticality levels AMCrtb is a dependable, comprehensive scheme.