An examination of self-similar network traffic and tactical radio experiments

Network traffic has been reported to exhibit self-similar properties over time. This has been demonstrated under numerous circumstances involving high speed networks using civilian protocol suites. Self-similarity has also been reported to arise as a result of different network processes. Different network protocols have been shown to affect the degree of self-similarity in network traffic. This is important because results indicate that the networks analyzed behave differently and exhibit different performance figures of merit depending on the traffic´s degree of self-similarity. This investigation focuses on very low bandwidth tactical radio nets with military protocols. The US Army Research Laboratory (ARL) has performed statistically controlled experiments to examine network performance using a combination of tactical communications protocols and tactical radios driven by traffic with an arrival rate to the network that is generated from a Poisson distribution. These channels are of extremely low throughput and use communications protocols different from the widely used civilian protocols. The assumptions and conditions of the reported sources of self-similar behavior in civil experiments are compared to the very low bandwidth conditions and the effects of the protocols used in the ARL experiments. These channels are of extremely small throughput by civilian standards, by factors of 5000 or more, and the tactical communications protocols are different from those based on the Transmission Control Protocol/Internet Protocol (TCP/IP) suite. This study examines the differences and attempts to draw inferences from them