Allocation and Optimisation of Mixed Criticality Cyclic Executives

Incorporating applications of differing levels of criticality onto the same platform in an efficient manner is a challenging problem. Highly critical applications require stringent verification and certification while lower criticality work may seek to make greater use of modern processing power with little to no requirement for verification. Much study into mixed criticality systems has considered this issue by taking scheduling paradigms designed to provide good platform utilisation at the expense of predictability and attempting to provide mechanisms that will allow for the verification of higher criticality work. In this thesis we take the alternative approach, we utilise a cyclic executive scheduler. Such schedulers are used extensively in industrial practice and provide very high levels of determinism making them a strong choice for applications with strict certification requirements. This work provides a platform which supports the highly critical work, alongside work of lower criticalities in a cyclic executive context. The aim being to provide a near-future platform which is able to support existing legacy highly critical software alongside newer less critical software which seeks to utilise multi-core architectures. One of the fundamental challenges of designing a system for a static scheduler is the allocation of applications/tasks to the cores and, in the case of cyclic executives, minor cycles of the system. Throughout this work we explore task allocation, we make extensive use of Linear Programming to model and allocate work. We suggest a limited task splitting technique to aid in system design and allocation. Finally, we propose two ways in which an allocation of work might be optimised to meet some design goal. This thesis proposes a scheduling policy for mixed criticality multi-core systems using a cyclic executive scheduler and explores the design, allocation and optimisation of such a system.