The EFTOS Voting Farm: A Software Tool for Fault Masking in Message Passing Parallel Environments

We present a set of C functions implementing a distributed software voting mechanism for EPX or similar message passing environments, and we place it within the EFTOS framework (Embedded Fault-Tolerant Supercomputing, ESPRIT-IV Project 21012) of software tools for enhancing the dependability of a user application. The described mechanism can be used for instance to implement restoring organs i.e., N-modular redundancy systems with N-replicated voters. We show that, besides structural design goals like fault transparency, this tool achieves replication transparency, a high degree of flexibility and ease-of-use, and good performance.