Scalable content distribution for social networking websites

Nowadays, much of content sharing happens through social networking websites where content diffusion exhibit certain patterns. Most of these contents are small in size but are very large in number. Since the edge/replica servers have limited storage capacity, therefore to minimize user perceived latency, we require selective replication techniques that efficiently decide which contents should be replicated where and when. We propose to use information dissemination patterns based on social networks to predict the possible locations where contents will become popular in future. This information can then be used to pro-actively place content replicas at the nearest possible edge servers of content demand.