A Framework for Analyzing Shared Resource Interference in a Multicore System

Integrated Modular Avionics (IMA) provides multiple benefits such as reduced SWAP-C and reduced recertification effort but requires deterministic behavior of the modular software that is integrated together. Lack of determinism can impact safety. The normal way to separate modular software to reduce unexpected effects is to use partitioning. The software in each partition is isolated from other partitions so that partitions cannot impact each other's required behavior. When partitions share computing resources, the system architecture must provide mechanisms to enforce determinism on that sharing and the flight certification artifacts must demonstrate the reliability of that enforcement. This paper proposes an approach for analyzing the use of shared resources by partitions in an IMA system, particularly for those built around a multicore processor, as a conceptual scheme to identify all possible avenues for interpartition interference that might cause non-determinism and then to validate and verify the system isolation mechanisms to bound such interference. First, we provide background by summarizing the IMA concept, reviewing basic computer architecture principles focused on shared resources, and defining isolation (also known as partitioning or separation). Second, we highlight the key literature on partitioning. Third, we provide a taxonomy of isolation techniques. Fourth, we propose the start of a framework for analyzing shared resource interference in light of the applied isolation techniques. The paper concludes with a summary of areas for further research.