Optimized cooperative streaming in wireless mesh networks

Abstract

Abstract-Peer-to-peer video streaming is a valuable technique to reduce the overhead produced by centralized and unicastbased video streaming. Key to the efficiency of a peer-topeer approach is the optimization of the logical distribution topology (the overlay with respect to the underlying network, the underlay). This work studies peer-to-peer streaming in wireless mesh networks for which the underlay is known. We propose an optimized, cross-layer approach to build the peer-to-peer distribution overlay minimizing the impact on the underlay. We design an optimal strategy, which is proven to be NP-complete, and thus not solvable with a distributed, light weight protocol. The optimal strategy is relaxed exploiting the knowledge of the betweenness centrality of the underlay nodes, obtaining two easily implementable solutions applicable to any link-state routing protocol. Simulation and emulation results (experimenting with real applications on a network emulated with the Mininet framework) support the theoretical findings, showing that the relaxed implementations are reasonably close to the optimal solution, and provide vast gains compared to the traditional overlay topology based on Erdös-Rényi models that a peer-topeer application would build.

Optimized P2P streaming for wireless distributed networks

“…This paper builds on a previous work in which we formalized the optimization problem and, introduced the relaxation strategy and evaluated its results [3]. This paper extends these results with further optimizations to make the proposal more efficient in overlays with a realistic topology, i.e.…”

“…The assumption we start from, which becomes a constraint of the optimization problem, is that each peer should receive exactly one copy of every chunk. In our previous work [3] we also assumed that each peer contributes to the dissemination serving chunks to other peers proportionally to its degree in O. This was motivated by the effort to keep the number of neighbors per peer constant (with a quasi-regular topology), so that the overall load was fairly distributed.…”

“…The work presented in [3] used two evaluation strategies: graph analysis with Matlab, and the Python Networkx graph library. Results were validated implementing the strategies on a real P2P video streaming platform, PeerStreamer [8].…”

“…In order to avoid a trivial solution we impose on each peer a minimum node degree d > log 2 (|P |), which also guarantees that the resulting overlay is connected with high probability 3 . This problem can be rephrased as: find the overlay graph Ω(F ) with minimum degree d defined by F = {S 1 , ..., S |P | } so that the norm of L(F ) is minimized subject to the following constraint…”

On the Computation of Centrality Metrics for Network Security in Mesh Networks

“…In fact, in real networks made of hundreds of nodes as few as 5 nodes are in the position to intercept about 90% of the potential traffic flows [3]. Besides security applications, centrality can help resources allocation [13], can be used to evaluate network robustness [14], to perform real-time dynamic allocation of services in a network [15], to optimize the configuration of link-state routing protocols [16] and to coordinate peer-to-peer protocols [17], [18].…”