Study of a linear coding enhanced swarm network

The present paper explores the behavior of a swarm network that relies on random linear codes to improve the content distribution to all the clients. The swarm implementation is simplified with respect to other known implementations (e.g. BitTorrent) by discarding the coordination mechanism between the peers, relying only on the linear encoding to still provide an efficient content dissemination inside the swarm. Two approaches are explored, one applies the linear coding only at the sources, while the second relies on network coding. The performances provided are assessed in several significant scenarios, by simulations in OMNET++, considering the average download time as the main metric, other metrics being also taken into consideration. The results show that the first approach is not viable, while the one based on network coding allows an efficient data dissemination inside the swarm, for various underlying network topologies. The second approach is simulated using also topologies which favor certain groups of clients due to clustering. The results show that the entire swarm network is positively influenced in such cases.