A practical Random Network Coding scheme for data distribution on peer-to-peer networks using rateless codes

In this paper we propose a practical Random Network Coding (RNC) scheme for data distribution in a peer-to-peer (P2P) overlay network. The use of RNC incurs a significant computational cost that, till present, has limited its deployment in practical applications. In this study it is shown that RNC complexity can be lowered by using Luby Transform (LT) codes to pre-encode the data and by letting intermediate nodes use RNC in a loworder Galois Field, i.e. GF(2). Moreover, we exploit a recently proposed variant of the Gaussian Elimination algorithm (OFG) to improve further both the creation of random combinations for RNC and the final decoding of the content. Our analysis is based on both analytical modeling and simulations over P2P overlay networks generated from random graphs and real snapshots of the PPLive streaming application. The results point out that using LT codes and RNC in GF(2) one is able to significantly improve the overall Email addresses: [email protected] (Valerio Bioglio), [email protected] (Marco Grangetto), [email protected] (Rossano Gaeta), [email protected] (Matteo Sereno) Preprint submitted to Elsevier September 3, 2012 performance in terms of both delay and bandwidth utilization at a reasonable computational cost. Finally, the RNC strategies we propose do not require any prior knowledge of the overlay network topology thus making them very general.