PeerChatter: A Peer-to-Peer Architecture for Data Distribution over Social Networks

Social networking services (SNSs) like Facebook, Twitter, Douban, etc. have become popular ways for data distribution over social networks. However, centralized SNSs have three main limitations: privacy, scalability and single point of failure. On the other hand, any distributed data distribution approach over social networks has to face the highly dynamic user behavior and support synchronous/asynchronous multicast. Focusing on those limitations and demands, we propose PeerChatter, a peer-to-peer architecture for data distribution over social networks. PeerChatter maintains an overlay network based on multi-level random graphs, called SkipCluster. The regular relation between levels makes SkipCluster support efficient routing, and the randomness inside levels ensure high robustness. Over SkipCluster, PeerChatter realizes the synchronous/asynchronous data distribution between friends and inside groups based on the publish/subscribe model. The experimental results show that PeerChatter can support efficient and reliable data distribution over social networks even in the highly dynamic network environment.