On the Use of Eigenvector Centrality for Cooperative Streaming

Abstract

The timely and efficient cooperative distribution of a streamlined content in a communication network is a key feature for many applications and services. One of the unsolved problems is the assignment of transmission rates to nodes given the constraints imposed by the topology, so that all nodes receive the stream with the minimal global use of resources. This paper addresses the problem exploiting the notion of eigenvector centrality. It shows that the problem can be solved efficiently in a distributed way if every node is aware of the full network topology and that in certain cases only local information on the network graph is sufficient.

On the Properties of Infective Flooding in Low-Duty-Cycle Networks

“…Let Θ j be the throughput (in terms of packets sent per second) that node j sustains in average and Θ the resulting column vector. Then, from the Theorem 1 in [6],…”

“…In the following subsections we combine our results on the optimality of receiver-equal resource allocation [6] together with an infective model dissemination on graphs to derive bounds on the performance of information flooding on mesh networks with minimal signaling.…”

“…As stated above, at steady state, the optimal resource allocation is the one that guarantees that every node receives the same amount of information (the same number of packets) per time-interval. The result presented in [6] can be summarized as follows; Section III derives delay bounds starting from this result.…”

“…Equation 2grants that every node has the same probability of receiving the information if we average over all possible sources s ∈ V . If A is column-uniform, which means that pakets are sent with uniform probability to the neighbors, these parameters can be computed by the nodes distributively by simply gossiping their neighbourhood set size [6].…”

“…The contribution of this paper stems from the encounter of a recent result published in the context of P2P streaming [6], with classical results from epidemic diffusion based on differential equations [7], [8]. We look at the broadcasting process as the "infective" propagation of data on a graph, that models the given topology of the network, hence the title, and we modify the infective capability of each node in such a way to optimize the information distribution.…”