ANI/STI: Measurements, Models, and Simulation Scenarios for Internet Research

International Computer Science Institute, Berkeley CA

Investigators

Abstract

Networking researchers work from mental models of the Internet's important properties. The scenarios used in simulations and experiments reveal aspects of these mental models, often including one or more of the following: (1) Flows live for a long time and transfer a lot of data. Simple topologies, like the "dumbbell" topology, with one congested link, are sufficient to study many traffic properties (2) Flows on the congested link share a small range of round-trip times (3) Most traffic across the link is one-way (4) Reverse-path traffic is rarely congested All of these models can deeply affect simulation and experimental results, and therefore evaluations of research, but none of them are confirmed by measurement studies, and some are actively wrong. Some divergences from reality are unimportant, in that they don't affect the validity of simulation results, and simple models help people understand the underlying dynamics of systems. However, as a community we do not yet understand which aspects of models affect fundamental system behavior, and which aspects can safely be ignored. This project proposes to build a much richer understanding of the range of realistic network models, and of the likely relevance of different model parameters to network performance. The lack of good measurements, lack of tools for evaluating measurement results and applying their results to models, and lack of diverse and well-understood simulation scenarios based on these models, are holding back the field. This program will directly address the models used by network researchers in simulations and experiments. Building on the wealth of existing measurement results and generating new empirical results where needed, this will create high-quality tools to facilitate ongoing measurement by this group and the wider network research community, so that the results keep up with the Internet's high rate of change. Synthesizing the relevant body of measurement results will provide concise intuitions, and a range of simulation-ready models, for use by researchers. An illustration of how each aspect of the models can influence simulations and experiments will be given. This work will not create a universal model of the global Internet (an immense and unnecessary undertaking). Instead, it will focus on models used by researchers in specific research areas, motivated by addressing their pressing modeling concerns. Focusing at first on the group's own areas of expertise, particularly modeling congestion-related mechanisms at a queue in a router, and later address other research areas, help others improve their measurements and measurement tools, along with the resulting models and simulation scenarios, and improve the relationships between experiments, models, and our shared understanding of simulation results.