A repeated game formulation of network embedded coding for multicast resilience in extreme conditions

Computer networks and data sharing applications are vital for our current society and fundamental for any available ICT solution, so that networking is considered as one of the key critical infrastructures and its correct behavior should be always enforced, even in case of disasters or severe execution conditions. Resilience is a strongly demanding nonfunctional requirement for current computer networks, and one of the key factors to provide it is represented by loss tolerance. From wired to wireless networks, exchanged packets can be lost due to routing anomalies or temporary malfunctioning at the networking hardware or software. The experienced loss pattern can be exacerbated by severe weather conditions in wireless environments and/or due to effects of a disaster. Proper recovery schemes must be put in place in order to provide loss tolerance and move a step forward realizing disaster-resilient networks. This work contributes to such an objective by proposing a loss-tolerant scheme based on forward error correction and properly implementing it in a distributed manner, by having multiple nodes generating coded information to be used for the reconstruction of the lost data without retransmissions. The problem of placing and tuning such coding nodes within a communication infrastructure has been tackled according to a repeated game theory formulation, and a preliminary evaluation demonstrates the achievable improvements in the delivery success rate with limited costs.