Multicore Resource Isolation for Deterministic, Resilient and Secure Concurrent Execution of Safety-Critical Applications

Multicores increasingly deploy spatial execution of safety-critical applications that demand a deterministic, resilient, and secure environment to meet the safety standards. However, multicores aggressively share hardware resources that leads to non-deterministic performance due to destructive interference from concurrent applications. Resource sharing not only hinders efficient resilient execution, but also introduces security vulnerabilities due to information leakage on side-channels. This work proposes a novel multicore framework that constructs isolated clusters of cores for each concurrent application. It guarantees concurrent applications with deterministic performance, as well as an efficient execution environment for resiliency and security. Moreover, the framework allows dynamic re-sizing of cluster sizes for load balanced execution of concurrent applications. However, it leads to diminished isolation between clusters, which opens various performance-resilience and performance-security tradeoffs.