Distributed Dataset Synchronization in Disruptive Networks
by Tianxiang Li, Zhaoning Kong, Spyridon Mastorakis, and Lixia Zhang
Disruptive network scenarios with ad hoc, intermittent connectivity and mobility create unique challenges to supporting distributed applications. In this paper, we propose Distributed Dataset Synchronization over disruptive Networks (DDSN), a protocol which provides resilient multi-party communication in adverse communication environments. DDSN is designed to work on top of the Named-Data Networking protocol and utilizes semantically named, and secured, packets to achieve distributed dataset synchronization through an asynchronous communication model. A unique design feature of DDSN is letting individual entities exchange their dataset states directly, instead of using some compressed form of the states. We have implemented a DDSN prototype and evaluated its performance through simulation experimentation under various packet loss rates. Our results show that, compared to an epidemic routing based data dissemination solution, DDSN achieves 33-56% lower data retrieval delays and 40-44% lower overheads, with up to 20% packet losses. When compared to the existing NDN dataset synchronization protocols, DDSN can lower the state and data synchronization delays from one-third to two-third, and lower the protocol overhead by up to one-third, with the performance difference becoming more pronounced as network loss rates go up.