Analysis of realized peer-to-peer streaming topologies by Kronecker graphs

Modeling the overlay graph of peer-to-peer (P2P) data dissemination is inherently difficult due to the high dynamics of the peer behavior and the high degree of connectivity if we regard a mesh-pull architecture. We present a solution of the latter issue using the mathematical tools of Kronecker graphs.We are able to capture accurately the static structure of the overlay graph arising from a P2P streaming application. To validate our proposal, we use a large data set of a worldwide distributed measurement campaign arising from the live streaming system SopCast. First, we determine the basic parameters of our Kronecker graphs model by an EM-algorithm. Then we use it to generate a synthetic graph and compare the properties of both structures. The results of our analysis illustrate the coincidence of several graph-theoretical properties of the realized and synthetic overlay graph and underline the great potential of a modeling approach based on Kronecker graphs.