Novel atomic action protocol for parallel systems with communication faults
Vincenzo De Florio, Geert Deconinck, M. Truyens, Rudy Lauwereins · 1998
Present-day embedded applications call for both parallel computing and fault tolerance (FT) . These requirements are driven by the increasing demand for processing power and the increasing economical cost related to system failure [1] . A typically required FT feature is an atomic action . The atomic action may be defined as the activity of a set of components where no information flows between that set and the rest of the system during that activity [2] . In general implementations hereof impose a number of non-realistic restrictions. Communication faults are excluded or blocking protocols are proposed [3,4] . This paper introduces an algorithm that provides a solution for atomic actions in a partially synchronous parallel environment with fallible communication The proposed solution opts for a time-bounded, fail-safe approach . We also present the implementation of the algorithm, which is part of a FT framework currently under development in ESPRIT project 21012, EFTOS (Embedded Fault TOlerant Supercomputing) [1] .