• Replication and availability in decentralised online social networks

      Hassan, Adil (University of Bedfordshire, 2017)
      During the last few years’ online social networks (OSNs) have become increasingly popular among all age groups and professions but this has raised a number of issues around users’ privacy and security. To address these issues a number of attempts have been made in the literature to create the next generation of OSNs built on decentralised architectures. Maintaining high data availability in decentralised OSNs is a challenging task as users themselves are responsible for keeping their profiles available either by staying online for longer periods of time or by choosing trusted peers that can keep their data available on their behalf. The major findings of this research include algorithmically determining the users’ availability and the minimum number of replicas required to achieve the same availability as all mirror nodes combined. The thesis also investigates how the users’ availability, replication degree and the update propagation delay changes as we alter the number of mirror nodes their online patterns, number of sessions and session duration. We found as we increase the number of mirror nodes the availability increases and becomes stable after a certain point which may vary from node to node as it directly depends on the node’s number of mirror nodes and their online patterns. Moreover, we also found the minimum number of replicas required to achieve the same availability as all mirror nodes combined and update propagation delay directly depends on mirror nodes’ number of sessions and session duration. Furthermore, we also found as we increase the number of sessions with reduced session lengths the update propagation delay between the mirror nodes starts to decrease. Thus resulting in spreading the updates faster as compared to mirror nodes with fewer sessions but of longer durations.