Mixed criticality scheduling in fault-tolerant distributed real-time systems

Abstract

Modern safety critical real-time systems are composed of tasks of mixed criticalities and the problem of scheduling them in a fault tolerant manner, on a distributed platform, is challenging. Fault tolerance is typically achieved by using redundancy techniques, most commonly in the form of temporal redundancy which involves executing an alternate task before the original deadline of the failed task. Additionally, studies like Zonal Hazard Analysis (ZHA) and Fault Hazard Analysis (FHA) may impose extra constraints on the re-executions, e.g., spatial separation of alternates, to improve reliability. In this paper, we present a method for scheduling mixed criticality real-time tasks on a distributed platform in a fault tolerant manner while taking into account the recommendations given by the reliability studies like ZHA and FHA. First, we use mathematical optimization to allocate tasks on the processors, and then derive fault tolerant and fault aware feasibility windows for the critical and non-critical tasks respectively. Finally, we derive scheduler specific task attributes like priorities for the fixed priority scheduler. Our method provides hard real-time fault tolerance guarantees for critical tasks while maximizing resource utilization for non-critical tasks.