On the consolidation of mixed criticalities applications on multicore architectures

Multicore architectures are very appealing as they offer the capability of integrating federated architectures, where multiple independent computing elements are devoted to specific tasks, into a single device, allowing significant mass and power savings. Often, the tasks in the federated architectures are responsible for mixed criticalities tasks, i.e. some of them are mission-/safety-critical real-time tasks, while others are non-critical tasks. When consolidating mixed criticalities tasks on multicore architectures, designers must guarantee that each core does not interfere with the others, introducing side effects not possible in federated architectures. In this paper we propose a hybrid solution based on a combination of known techniques: lightweight hardware redundancy, implemented using smart watchdogs and voter logic, cooperates with software redundancy, implemented using software temporal triple module redundancy for those tasks with low criticality and no real-time requirement, and software triple module redundancy for tasks with high criticality and real-time requirement. To guarantee lack of interference, a hypervisor is used to segregate the execution of each task in a dedicated resource partition. Preliminary experimental results are reported on a prototypical vision-based navigation system