A framework for adaptive reconfigurable space-borne computing platforms for run-time self-recovery from transient and permanent hardware faults

摘要

Mitigation of radiation effects is one of the major problems for space-borne computing platforms. The presented work proposes an approach for building reliable, hardware fault adaptive stream processing platforms for space applications. The proposed concept is based on architecture-to-fault adaptation by run-time hardware reconfiguration. The concept assumes representation of system components in virtual form (configuration bit-streams for partially reconfigurable FPGAs) and dynamic allocation and/or relocation of components in predetermined slots in the target FPGA. In case of transient faults (e.g. SEU) the affected component can be reloaded to the same slot. In the case of permanent faults the affected component can be relocated to a reserved slot. Experimental results have shown that both restoration procedures can be done in runtime. The work describes novel procedures for run-time on-chip self-assembling/repair procedures of an application specific processors (ASP). The self-synchronization aspect of the above ASP also is discussed in detail. The feasibility of the proposed approach has been proved by testing all above procedures on a prototype platform based on a Xilinx Virtex-4 FPGA. The timing and hardware overhead for the required framework infrastructure have also been analyzed and discussed.