Stochastic dataflow graph models for the reliability analysis of communication networks and computer systems

The literature is abundant with combinatorial reliability analysis of communication networks and fault-tolerant computer systems. However, it is very difficult to formulate reliability indexes such as survivability, repairability, probability of successful communication (or job completion) of communication networks using combinatorial methods. These limitations have led to the development of time-dependent reliability analysis using stochastic processes.;Traditional stochastic-process reliability models use random variables to describe only the failure phenomenon of the system. They do not address how the failure behavior is influenced by job completion rates (or service rates) of the corresponding processors. Neither do they provide a formalized methodology for translating the system into Markov process, and thus the verification of the correspondence of the model to the actual system is not straightforward.;In this research, time-dependent reliability analysis techniques using Dataflow Graphs (DFG) are developed. The chief advantages of DFG models over other models are their compactness, structural correspondence with the systems, and general amenability to direct interpretation. This makes the verification of the correspondence of the dataflow graph representation to the actual system possible. Several DFG models are developed and used to analyze the reliability of communication networks and computer systems. Specifically, Stochastic Dataflow graphs (SDFG), both the discrete time and the continuous time models, are developed and used to compute time-dependent reliability of communication networks and computer systems. The repair and coverage phenomenon of communication networks is also analyzed using SDFG models.