State Management in Large Scale Group Communication

Current state management practices heavily rely on centralized architectures that pose scalability bottleneck and are expensive to adopt and deploy. In this paper, large scale state management issues are considered and addressed from design perspectives. The intra zonal communication strategies presented in [2][5] assume that peer knows its position in the global coordinate system. In this work, we drop that assumption and approximate peer position through the relative end-to-end delay with respect to landmarks. The technical analysis presented in [4] shows that the protocol has higher diameter in terms of hop distance compared to NICE [6], although it has lower diameter in terms of end-toend delay among the peers. To address that problem we present multiphase mesh construction technique that requires further partitioning of each region into several sub-regions. Both server-side and client-side resources are effectively utilized in collaborative manner while respecting their roles in the system.