EDCC 2005, LNCS 3463, pp. 230 – 245, 2005. © Springer-Verlag Berlin Heidelberg 2005 Fast Run-Time Reconfiguration for SEU Injection David de Andrés, José Albaladejo, Lenin Lemus, and Pedro Gil Fault Tolerant Systems Research Group, Universidad Politécnica de Valencia, DISCA-ETS Informática Aplicada, Campus de Vera #14, E-46021, Valencia, Spain {ddandres, jalba, lemus, pgil}@disca.upv.es Abstract. Mission critical systems are computer-based control systems that people depend on, often for their livelihoods, sometimes for their lives. As the integration scale of VLSI increases, computer-based systems are more prone to be affected by Single Event Upsets (SEUs). FPGAs have proven to be very use- ful devices for the injection of SEUs in order to assess the dependability of those systems. The conventional approach for injecting SEUs following the Run-Time Reconfiguration methodology involves reconfiguration times de- pendent on the complexity of the model. In case of complex models, it will greatly increase the execution time of fault injection experiments. This paper presents a new approach for the injection of SEUs into the memory elements of the system under test by means of FPGAs. This approach takes benefit of the FPGA architecture to minimise the reconfiguration time for SEU injection, ob- taining an speed-up of near two orders of magnitude. 1 Introduction In recent years, there has been a great increase in the use of computer-based systems in mission critical systems, such as automotive systems, aircrafts, telecommunication, etc. The occurrence of a failure in this kind of systems can cause the loss of huge amounts of money or even worst the loss of human lives. Therefore, there has been a great interest in developing different techniques to validate the fault tolerance capa- bilities of these systems and evaluate their dependability. At the same time, the integration level of electronic circuits is experiencing a con- tinuous increase. The complexity of these new systems is making it difficult to guar- antee a certain level of dependability due to the increasing rates of occurrence of tran- sient (also known as soft errors or single event upsets (SEU)), intermittent and permanent faults [1]. Fault injection appears as a suitable means to evaluate the dependability of this kind of systems. The main aim of fault injection is to enable the analysis of the behav- iour of the systems in the presence of faults. The process of introducing such faults into the system in a controlled way is named injection. It is well known that the analy- sis of the results obtained from these experiments is useful to assess the dependability of the system under study. M. Dal Cin, M. Kaâniche, and A. Pataricza (Eds.):