PRACTICABLE SOFTWARE FAULT TOLERANCE

A METHODOLOGY IS PROPOSED FOR TARGETING SPECIFIC ERROR DETECTION AND RECOVERY TECHNIQUES AT FREQUENT AND SEVERE CAUSES OF SOFTWARE FAILURES. THE SPECIFIC TECHNIQUES ARE MUCH LESS COSTLY TO IMPLEMENT THAN COMPREHENSIVE MULTI-VERSION APPROACHES THAT ARE WIDELY DESCRIBED IN THE TECHNICAL LITERATURE. A KEY ASSUMPTION IN OUR APPROACH IS THAT MOST FAILURES IN OPERATION TACTICAL SYSTEMS ARE DUE TO RARELY ENCOUNTERED DATASETS OR COMPUTER STATES, AND THAT NORMAL OPERATION CAN BE RESUMED IF THE RESULTS OF FAULTY OPERATIONS CAN BE CLEARED FROM THE SYSTEM. TO AID LESS EXPERIENCED PERSONNEL IN SELECTING AND PLACING THE FAULT TOLERANCE PROVISIONS, AN EXPERT SYSTEM WILL BE DEVELOPED.