An empirical comparison of software fault tolerance and fault elimination

A large-scale experiment comparing software fault tolerance and software fault elimination as approaches to improving software reliability is described. Results are examined that apply to the appropriateness and underlying assumption of the two i.e., reducing standard testing procedures when using voting to achieve fault-tolerance in operational software and using voting in the testing process. Among other results, it was found that n-version programming did not tolerate most of the faults detected by the fault elimination techniques. The results also cast doubt on the effectiveness of using voting as a test oracle.<<ETX>>