CAPRES: a software tool for modeling and analysis of fault-tolerant computer architectures

An engineering software tool is described that was developed to support efficient and cost-effective design of large-scale, parallel, fault-tolerant computer architectures (FTCAs). The tool, termed CAPRES (computer-aided performance and reliability evaluation system), allows the user to predict, by analytical means, the performance, reliability, and performability of candidate computer architectures executing candidate workloads. The algorithms of CAPRES use multidisciplinary techniques from hierarchical queuing-network theory, semi-Markov processes, hybrid-state systems theory and numerical integration. The sophisticated user interface and high-level front-end of CAPRES make this software system a truly integrated package for FTCA analysis and evaluation. To illustrate the role of CAPRES as a tool to aid in the analysis of FTCAs, an example of an Encore Multimax multiprocessor executing a parallel weapon-target-assignment algorithm is presented.<
>