Setting checkpoints in legacy code to improve fault-tolerance

摘要

In this paper we describe the results of a study of the insertion of checkpoints within a legacy software system in the aerospace domain. The purpose of the checkpoints was to improve program fault-tolerance during program execution by rolling back system control to a saved state from which program execution can continue. The study used novice programmers for the determination of where the checkpoints were to be added. The focus was on the programmer's understanding of the code, since this affected how the checkpoints were placed. The results should provide guidance to those interested in improving the fault-tolerance of legacy software systems, especially those written in older, nearly obsolescent programming languages.