This paper [1] addresses some dominate issues in Internet simulation. The authors believe that the immense challenge in simulating the global Internet due to the great heterogeneity and rapid change.

First, the authors examine the properties of Internet, especially those aspects that can affect the simulation. The diverse underlying technologies and administrative entities, the huge size and its unpredictable growing rate, and the immense variations over time in which applications are used and in what fashion, all produce big difficulties for us to simulate the Internet. In addition, the invisible and hard to construct topology, various link properties, and asymmetric routes, plus the different implementations of protocols make the situation much worse.

Then the authors argue that the ¡°traffic shaping¡± using trace data cannot properly characterize the network property. Hence, they suggest source-level simulation instead of packet-level simulation driven by trace.

They also imagine the possible future properties of Internet, including pricing mechanism, routing architecture, multicast, QoS, caching and prevalence of VR games.

The two main methodologies to cope these simulation difficulties are finding invariants and exploring the parameter space. The promising candidates on the first one include ¡°self-similar¡± packet arrivals, Possion-modeled user session arrivals, log-normal distribution of connection durations and heavy-tailed file sizes, etc. The authors encourage exploring parameter space by suggesting that more tries on parameter space, more chances to get ¡°simulation scenario invariants¡±.

Followed by the introduction on their VINT project, which aims at the multi-protocol simulation at large scale, the authors discuss the role of simulation. They argue that the simulation, experimentation, measurement and analysis are all key roles and complementary for each other.

Generally, this paper can serves as a classic research report in this area. It not only summarizes the main problems in the simulation of Internet, but also addresses the potential approaches which are all comprehensive. Moreover, the predicates of the future properties of Internet show the insight and creativity. Although some of these predicates turn out being not the case today, like the pricing structures, the overall contribution is significant. This paper will help the researchers to shape a thoughtful methodology and lead the research to a promising direction.

[1] Vern Paxson and Sally Floyd. Why We Don't Know How to Simulate the Internet. Proceedings of the Winter Simulation Conference, December 1997.