On the cost of low-fidelity modeling: The case for a performability approach

Abstract

The complexity of modern systems often precludes the development of exact models for reliability measures. Thus, a common issue is the achievement of fidelity in reliability models in light of system complexity. This issue grows with the stringency of reliability requirements imposed on the system. How to conduct a reliability analysis so that adequate fidelity is provided while maintaining manageable complexity is the topic addressed here. The performability approach, which takes performance effects into consideration in handling the complexity in reliability analysis, is proposed as an important element of the process. A case study on network reliability analysis is presented to demonstrate the approach. The example shows that the performability approach leads to cost-effective designs, while low-fidelity, overly conservative approaches lead to over-design of the system or overlooked effects of system operations. This approach has been applied to a real-life satellite navigation system, where it proved to be a useful technique in conducting reliability analysis for complex systems.