In-Hardware Fault-Tolerance Controller for Multi-FPGA Clustered Architectures
Abstract
The growing adoption of Commercial-Off-the-Shelf (COTS) components in space applications necessitates addressing performance and reliability challenges in radiation-intensive environments. Field-Programmable Gate Arrays (FPGAs), widely employed in these systems, offer exceptional computational capabilities but are vulnerable to radiation-induced faults, such as Single Event Upsets (SEUs), which can cause unpredictable behavior or even system failure. This paper presents a solution for enhancing the reliability of FPGA cluster-based systems through two custom-designed hardware solutions, the beacon controller and the robust bus, as well as error correction and workload redirection mechanisms. They have been integrated into an FPGA computing cluster for space telecommunication applications and tested against radiation-induced effects by fault injection techniques and radiation test campaigns. Experimental results demonstrated that the proposed techniques significantly reduce the system downtime, achieving up to 99% system availability, and reduce the radiation-induced failure rate effects below 4%. These results validate the proposed mitigation strategies as a robust and reliable approach for ensuring system dependability in harsh environments, demonstrating excellent performance through a fully FPGA-based design that eliminates the need for external controllers, unlike most state-of-the-art solutions.