Toward realistic network topologies

Obtaining realistic network topologies is crucial when designing routing protocols or trying to study the structure of the internet. They are usually not freely available for various reasons. In this context, we present Hydrogen, a new engine capable to infer network topologies from external measurements. Our aim is to provide the scientific community with a new tool capable of generating a router-level topology with link weights. This tool source code is freely available at https://bitbucket.org/gfeuillen/hydrogen/overview. In this thesis, we first explain how to infer router topologies, we then describe the link weights inference and then provide the topologies obtained by combining the two firsts parts. Throughout this thesis we provide quality measures of our implementation in order to prove its efficiency or find its weaknesses. More specifically, these measures allow us to prove that our topologies sometimes present inaccuracies (mostly due to shortcomings in the external measurements) while the inference of link weights is able to produce weights correctly characterizing the routing.