Credibility and Validation of Simulation Models for Tactical IP Networks

The task of TNO is to provide predictions of the scalability and performance of the new all-IP tactical networks of the Royal Netherlands Army (RNLA) that are likely to be fielded. The inherent properties of fielded tactical networks, such as low bandwidth and quality of service (QoS) policies require models of the key tactical network devices that, under similar traffic conditions, accurately simulate the behavior of the devices they represent. TNO uses OPNET for modeling military communication networks. However, validation reports of the standard models of the key devices provided by OPNET, or experimental values to compare their behavior with the behavior of the devices they represent, could not be obtained from the device vendors or OPNET. Subsequently, TNO embarked on experiments to validate models of the key IP routers and switches. The first results from this ongoing study were presented at the OPNET user conference in 2002. These results, and the ones in here, reveal the validity of the standard models. The experiments provide effective values for basic model attributes such as packet forwarding speed, effective queue sizes, and QoS mechanisms. Experiments revealed that effective values can differ significantly from vendor specifications. Sometimes the attributes of the corresponding standard OPNET model could not be tuned to match its behavior sufficiently with the measured one. Modifications were implemented to improve the match when the effort was not too great. Some flaws were considered too fundamental for modification but the modeler is now at least aware of them. This paper reports new experiments, simulations and results in our effort to extend our suite of valid models for tactical equipment. To incite discussion and to place our work in perspective, this paper also gives a concise overview of related studies reported in the literature on the validity of models and credibility of simulation studies of communication networks