Assessing system performance using component level performance specifications

Performance of software components may be characterized in a way that permits software architects to predict response times that result after integrating multiple components. Using information about individual component execution time and invocation rates, this method predicts processor utilization and “thread” latency (where a thread is an executed string of components). The method derives component budgets which can be individually verified via empirical tests and which assure system response times meet specified requirements. In the event budgets do not support satisfactory system response times, the method determines what components should be optimized in order to produce the desired system result. When calculating budgets or when identifying optimization goals for components, the method considers difficulty of component optimization. The method is based on a simple application of mathematical concepts from queuing theory and optimization theory and may be implemented using a spreadsheet. This method was used to develop component based budgets for a large complex software intensive system