Methodological considerations for research with network simulators

Simulation allows one to recreate scenarios for studying, through the use of computer programs, without needing to implement them in the real world. This reduces the cost of experimentation and increases the level of control one has over the system model. In the context of computer networking research, simulators are very useful for the analysis of network protocols and components. Analysis of TCP performance is one of the most studied problems. The original implementation of TCP used to contemplate packet loss due to network congestion only, typical of wired connections. The emergence of wireless networks called the attention to some faults in TCP, mainly its inability to distinguish between the many sources of packet loss in this kind of connections: out-of-range devices, mobility, signal interference, etc. There are many authors that talk about the advantages and disadvantages of system modeling with a simulation approach, the structure for courses, and its application in certain domains. There's also a vast literature on the topic of protocol analysis using network simulators, NS2 in particular. This paper attempts to integrate knowledge of computer networking and telecommunications, with simulation tools capabilities and expertise in TCP performance evaluation research, in an attempt to explain some useful methodological considerations in such enterprise.