A NEW SIMULATION METHOD FOR SYSTEM DEPENDABILITY ANALYSIS

TODAY'S COMPLEX SYSTEM ENVIRONMENT AND COMPETITIVE ECONOMIC ENVIRONMENT DEMAND PRECISE ANSWERS TO QUESTIONS CONCERNING DEPENDABILITY AND QUALITY. CURRENT TOOLS ARE NOT PROVIDING THE ABILITY TO MODEL COMPLEX FAULT HANDLING SCENARIOS AND OFTEN REQUIRE SIMPLIFYING ASSUMPTIONS ABOUT FAILURE RATES, REPAIR RATES, AND INITIAL STATES WHICH LEAD TO INCORRECT SOLUTIONS. ACCURATE DEPENDABILITY EVALUATIONS ARE A CONCERN OF BOTH MANUFACTURERS AND USERS OF SYSTEMS WHO OFTEN INVEST MILLIONS OF DOLLARS TO KEEP SYSTEMS OPERATING CORRECTLY. A DEPENDABILITY TOOL THAT CAN ADDRESS THESE CONCERNS IS IN DEMAND IN BOTH THE COMMERCIAL AND MILITARY ENVIRONMENTS. WE PROPOSE A SIMULATION TOOL BASED ON RAINBOW NETS WHICH WERE DEVELOPED BY ALLEN JOHNSON. RAINBOW NETS OVERCOME LIMITATIONS OF PREVIOUS NET-BASED MODELS BY SIGNIFICANTLY REDUCING THE MODEL COMPLEXITY. THISREDUCTION IS ACHIEVED BY HIERARCHICAL TOKENS, WHICH CAN BE CREATED, DESTROYED, AND HAVE IDENTITIES AND FEATURES,AND BY TRANSITIONS WHOSE ACTIVATION IS GOVERNED BY A SET OF RULES OR AN ALGORITHM. TO DEMONSTRATE FEASIBILITY, THIS RESEARCH WILL QUANTIFY IMPROVEMENTS GAINED OVER EXISTING TECHNIQUES AND PROVIDE SEVERAL NEW DEPENDABILITY MEASURES. IN ADDITION, SPECIFICATIONS WILL BE WRITTEN FOR A COMMERCIALLY VIABLE SOFTWARE TOOL. IN GENERAL, THIS WORK WILL LEAD TO A GREATER UNDERSTANDING OF HOW FAULTS AFFECT COMPUTER SYSTEMS.