Performance extrapolation that uses industry benchmarks with performance models

Predicting application performance for any hardware change is challenging as multiple factors like environment, application architecture, workload etc also need to be considered. Industry benchmarks that provide means to compare hardware performance however lack in giving insight into the applications' service levels. LQN models help in performance analysis of multi-tiered, distributed applications; but lack in predicting impact of vertical scaling. So to predict an application's performance for hardware changes, a hybrid methodology of these two approaches can be used as described here. The approach is illustrated and its results are validated for a standard J2EE application in this paper.