Balanced content replication in peer-to-peer online social networks

This paper presents an effective content replication scheme for peer-to-peer online social networks (P2P-OSN). The topology of P2P-OSN is defined by the social network of the participants. P2P-OSN allow people to share content and run applications with their 1-hop friends in decentralized fashion, while denying access to their potentially private data to other users. Content replication in these networks is difficult because users can place replicas only at their 1-hop friends, and this could substantially skew the storage availability in the network. Our main contribution is a distributed replication method that prevents the skewness of available replication storage across the network and improves replication fairness/success without relying on global knowledge of the social network. We developed a new centrality metric, EasyRank, which is calculated at each peer and finds the underlying connectivity structure responsible for introducing the skewness of storage availability. Our distributed replica placement algorithm places the replicas after ranking the potential storage peers using their EasyRank scores and currently available storage. We evaluated our solution with social graphs from Facebook and Google+ having more than 4,900 vertices and 720K edges. The evaluation is done for both stable and emerging social networks. The results show that EasyRank-based replication achieves the fairest storage allocation and maintains the most balanced storage availability among the tested methods. Thus, it provides the highest replication success rate.