Energy and reliability management are important design constraints for real-time embedded systems. We consider the problem of achieving a given reliability target for a set of periodic real-time applications running on a multi-core system with minimum energy consumption. Specifically, we observe that the emerging multicore platforms provide ample opportunities to use task replication to achieve reliability targets and mitigate the negative impact of Dynamic Voltage Scaling (DVS) on the rate of transient faults leading to soft errors. However, while it allows using lower execution frequencies, replication may also increase overall energy consumption due to additional task copies. Our objective is to determine the level of replication and frequency assignment for each task, as well as task-to-core allocations, in such a way to achieve the target reliability levels with minimum energy consumption. We first identify the subtle interplay be- tween the processing frequency, replication level, reliability, and energy consumption on DVS-enabled multicore systems. Then we show that the problem is intractable in the general case and propose our energy-efficient replication (EER) algorithm as an approximate solution. We also show how the framework can be extended to tolerate a given number of permanent faults affecting processing cores. We evaluate the performance of our proposed scheme through extensive simulations. The simulation results indicate that through our algorithm, a very broad spectrum of reliability targets can be achieved with minimum energy consumption through the judicious use of replica and frequency assignment.
展开▼
