Exploitable results
Parallel HPC (High Performance Computing) is being increasingly used as a part of complex electromechanical systems, using embedded computing for process control, pattern recognition, fast data and signal processing. In these applications, system failures are connected with high costs, or even worse, with safety hazards. Hence an adaptable, scaleable and cost-effective fault tolerance concept is needed to reach system behaviour like graceful degradation or safe-stop on the occurrence of faults.
EFTOS (Embedded Fault TOlerant Supercomputing) provides a framework for the implementation of fault tolerance in embedded supercomputing applications. This framework, with powerful, adaptable and standard solutions, provides programmers with a flexible approach that makes such applications more dependable. This approach will shorten the time to market, facilitate software development, maintenance and upgrading, and reduce dependence on specific hardware.
The specifications for the framework have been deduced from live industrial applications like mail sorting (Siemens) and control of high-voltage substations (ENEL). These applications clearly illustrated the lack of a standard approach towards fault tolerance and the common choice for non-reusable ad-hoc solutions. Hence the EFTOS framework was setup with extendibility and customisability in mind, as is needed for any long-life software.
Project URL : http://www.esat.kuleuven.ac.be/acca/ft/eftos
Buscando datos de OpenAIRE...
Se ha producido un error en la búsqueda de datos de OpenAIRE
No hay resultados disponibles