Empirical evaluation of protocol performance over tactical networks

The authors describe an experiment to evaluate the performance of a tactical network. The purpose of the experiment was to quantify the effect of three parameters, message length, message arrival rate, and transmission mode, on throughput and delay. The experiment was designed to allow an analysis of variance of these factors to examine their significance. Further data summaries helped quantify the differences these factors made in network performance. The results of this experiment provide statistically sound baseline information which can be used as input for network simulations, as guidelines for designing communications architectures and protocols, and for future experiments on combat radio nets. For throughput, the analysis of variance showed that message length was a more significant factor than arrival rate; their interaction was negligible. The transmission mode was not a statistically significant factor. For delay, message length compared to arrival rate was the more significant factor for both single-channel and frequency-hopping data.<
>