System development risk reduction using a system performance model

Abstract

A major risk to system development lies in the improper determination of real time software performance requirements early in the development cycle. All too often, performance concerns are put off until test and integration, only to discover the software or hardware developed cannot perform the needed functions within the time allotted. One good way to avoid this embarrassing and costly problem is to develop a System Performance Model (SPM). A System Performance Model is a discrete event system simulation that measures system event timing, resource usage, and access scheduling interactions. SPM outputs include measured resource, executive, and input/output usage with detailed thread execution profiles. The SPM enables design engineers to focus on timing and performance issues as the requirements are assessed. As the preliminary design is partitioned between hardware and software, the SPM provides full timing and performance assessment data, enabling designers to establish achievable timing budgets for all system components. The SPM is tested against operational scenarios to derive a full parametric and sensitivity response for the proposed design. Armed with solid engineering data, designers proceed to detailed design with full knowledge of the critical hardware and software components to be developed. The established time budgets and sensitivities enable software managers to know exactly which modules and tasks require extra care, greatly reducing the risks. The SPM is used throughout the entire system development life cycle, identifying and solving timing and performance problems before significant investment is made in hardware and software modifications. This paper describes how the SPM is used throughout the system life cycle. Key performance and timing analysis topics are presented, along with an example showing the subtle dynamic interaction problems encountered by real time embedded software.