|Thesis abstract: |
The PhD thesis proposes a methodology for the design of reliable embedded systems on multi-FPGA platforms. The objective is the exploitation of commercial SRAM-based FPGAs in mission-critical scenarios. Embedded systems and applications for space missions, where maintenance and system¿s lifetime are critical aspects, constitute the driving application scenario of the thesis, that has been partially supported by European Space Agency.
The idea is to achieve fault tolerance by exploiting the reconfigurable properties of the devices. We identify two categories of faults based on the possibility to recover from them by reconfiguration; recoverable and non-recoverable faults. Recoverable faults can be mitigated by reconfiguring the system, and possibly only the faulty sub-system portion, with the same configuration used before fault occurrence. Non-recoverable faults permanently compromise part or all of the device, such that further use of the corrupted portion of the device must be avoided and the logic hosted must be moved in a different location.
The proposed reliability-aware methodology realizes autonomous fault tolerant systems implemented on multi-FPGA platforms. The obtained system can continue working even if faults occur, thus increasing both reliability and lifetime.