Experience with using simulations for congestion control research

As part of the CESNET End-to-end performance project [1] we investigate enhancement of congestion control mechanisms providing better performance on fast long-distance networks. Network simulations end emulations are techniques that enable us to compare different alternatives and evaluate their performance under varying and well-defined network conditions, which would not be possible on real networks with unpredictable traffic dynamics. In this report we summarize our experience with using well known ns2 network simulator for studying congestion control mechanisms. We discuss difficulties that we encountered and describe our own enhancements to ns2, enabling to study more alternative and obtain more detailed information about their behaviour. 1 What is network simulation and how can it be useful? Different researcher use somewhat different definitions of simulation depending on the way they concieve it and use it. In the context of computer networks, we can say that simulation imitates behaviour of a real world or hypothetical future system by a computer program for the purpose of study. Properties of simulation include the following: • Simulation creates a model of a system, which is used to explain system behavior and to see how the system performs under varying conditions to design the system with desirable performance characteristics • Simulation usually represents the system on a smaller scale for easier study when compared to the full scale • Simulation allows to study a system in well-defined and well-known conditions, repeatability is necessary in order to understand events