This paper presents a concept for metric-aware dynamic fidelity modeling (MADFM), a simulation framework capable of intelligently applying fidelity consistent with analysis goals. (Fidelity refers to the accuracy of a simulated representation when compared to the real world.) As part of efforts to develop this concept, JHU/APL has designed and implemented a multiple-fidelity simulation framework that relies on software agents to interface diverse analysis environments. We use our prototype to conduct a sample analysis, in which actions occurring in a user behavior model trigger network traffic in an application model that in turn affects traffic rates in a network simulation. We discuss our system design and present results that demonstrate the validity of the prototyped framework. Our ongoing research seeks to enhance this prototype by adding advanced capabilities for requirements analysis and dynamic model selection.
Unfortunately, ACM prohibits us from displaying non-influential references for this paper.
To see the full reference list, please visit http://dl.acm.org/citation.cfm?id=1404692.