The scaling hypothesis: Simplifying the prediction of network performance using scaled-down simulations

As the Internet grows, so do the complexity and computational requirements of network simulations. This leads either to unrealistic, or to prohibitely expensive simulation experiments. We explore a way to side-step this problem, by combining simulation with sampling and analysis. Our hypothesis is this: if we take a sample of the traffic, and feed it into a suitably scaled version of the system, we can extrapolate from the performance of the scaled system to that of the original. We find that when we scale a network which is shared by TCP-like flows, and which is controlled by a variety of active queue management schemes, then performance measures such as queueing delay and the distribution of flow transfer times are left virtually unchanged. Hence, the computational requirements of network simulations and the cost of experiments can decrease dramatically.