FRESH: Fault-tolerant Real-time Scheduler for Heterogeneous multiprocessor platforms

Abstract

Real-time embedded systems are designed to execute precise functions within strict time constraints, utilizing microcontrollers, memory, and input/output devices. These systems’ critical component are the scheduler, responsible for efficient resource allocation and job scheduling based on priority and available resources. Multiprocessor platforms have been adopted to enhance performance, scalability, redundancy, and flexibility, employing diverse scheduling approaches. Fault tolerance is crucial in safety-sensitive systems that operate in real-time as they offer advantages by dynamically adapting to temporary faults, thereby ensuring system reliability and meeting performance requirements without sacrificing resource efficiency. Additionally, reducing dynamic energy consumption plays a vital role in improving battery life and reliability and adhering to power constraints in applications. Existing fault-tolerant schemes primarily focus on homogeneous multiprocessor systems or dual-type heterogeneous systems. This work introduces a novel heuristic scheduler named FRESH, which effectively addresses energy management and fault tolerance challenges in systems with various processor types. To validate the proposed approach, we conduct experiments using benchmark programs which show that FRESH is able to create a high number of secondary copies for the jobs to mitigate transient faults and also reduce significant energy consumption.