FEST: Fault-Tolerant Energy-Aware Scheduling on Two-Core Heterogeneous Platform
Piyoosh Purushothaman Nair, Rajesh Devaraj, Arnab Sarkar · 2018
Energy awareness and fault-tolerance have emerged as important design constraints in the development of modern safety-critical real-time systems. Devising scheduling strategies for such systems implemented on heterogeneous platforms is a challenging as well as a computationally demanding problem. As a consequence, today we face a scarcity of low-overhead scheduling techniques which are applicable to heterogeneous platforms. This work attempts to develop an efficient energy-aware heuristic scheme called, FEST, for the fault-tolerant scheduling of real-time tasks on a two-core heterogeneous platform, where one core is high-performance and the other core is power-efficient. In order to provide fault-tolerance against transient processor faults, we consider a standby-sparing approach where the power-efficient core is used to execute primary task versions while the high-performance core is operated as a spare to re-execute fault affected tasks (i.e., backups). Since, the execution of backups scheduled on the spare core will be cancelled subsequent to the fault-free execution of their primaries, we employ the Dynamic Power Management (DPM) technique on both cores to minimize energy consumption. Further, FEST utilizes backup-backup overloading to minimize energy consumption while guaranteeing tolerance against a given number of transient processor faults. Experimental results reveal that our scheme achieves good energy savings as compared to the state-of-the-art.