A system engineering methodology for predicting scalability of very large real-time databases

Designing scalable systems to support real-time access of large databases requires extraordinary methods for accurate performance analysis and for evaluating implementation alternatives. The paper describes a system engineering methodology which has been used at Harris to model and demonstrate such a system. The methodology consists of simulation results that are iteratively applied with a physical benchmark configuration of subset functionality until identical results are achieved. The simulation model is then scaled to represent the desired database configuration. Execution of the upscale model produces reliable results when tuned with performance data extracted from the iterative process described above. This methodology represents a thorough and believable approach to deriving key performance parameters on large systems for which there are no reliable data