Identification and analysis of simple models for the response time of complex computer systems

Graduation date: 1980A simple approach to developing approximate analytical\ud performance models for complex systems (such as distributed\ud processing systems or computer networks) is presented.\ud The approach is applicable in cases where measurement\ud or simulation data describing system performance are\ud available but the system is too complex for exact analytical\ud models to be quickly developed.\ud Known results from queueing theory are first used to\ud indicate the appropriate functional form of analytical\ud formulas for system performance, then modified regression\ud techniques are used to estimate unknown parameters in these\ud formulas. In all cases examined in this thesis, this\ud simple approach has closely predicted the performance of\ud the system studied. It appears to be applicable to a wide\ud variety of systems, and is especially valuable when it is\ud combined with further analysis of the system to analytically\ud explain the parameters obtained by regression