Viral Spread in Mobile Social Network using Network Coding

Abstract

Delay tolerant ad hoc network between mobile devices is a promising paradigm which can avoid significant cost of cellular data transfer. Many interesting applications such as peer-to-peer file transfer, micro-blogging etc. are possible on this network. We propose viral spread (Many-to-all broadcast) for such networks. Viral spread can be useful to variety of such applications as well as for routing protocols which require flooding of control messages. As these devices are carried by humans, we exploit properties of human mobility like overlapping community structure and heterogeneous popularity of nodes for efficient forwarding. Buffering all packets from all sources of community at each node is prohibitive. As buffer space is limited and buffer occupancy level also has an impact on energy requirements, to reduce buffer usage, we probabilistically buffer received packet for forwarding. To offset for the loss in performance due to this, we employ network coding. Network coding is a mechanism in which nodes encode two or more incoming packets and forward encoded packets instead of forwarding them as it is. To compare performance of viral spread, we also modify epidemic routing from unicast to broadcast and from deterministic buffering all the packets to probabilistic buffering. As simulation results suggest, viral spread achieves significantly high packet delivery percentage and low average delivery delay than modified epidemic routing. From simulation results, we also suggest optimal values of forwarding probability and generation size to be used in viral spread.