Intelligence Serviced Task-driven Network Architecture

Internet carries a variety of emerging applications which exhibit diverse requirements on transmission. The traditional end-to-end transmission scheme that provides only location-based content delivery has become the bottleneck of deploying these emerging applications. In recent years, newly proposed network addressing schemes enable diverse transmission modes. But each of the schemes is advantageous only in its specific scenario. Integrating all these addressing schemes to optimize transmission efficiency for all applications is still an open problem.To this end, we present an intelligence serviced task-driven network architecture. It applies federated learning to identify the transmission characteristics of different tasks deployed. Then, it assigns appropriate addressing schemes and thus all the tasks of each application take advantages of the new addressing schemes. Our prototype implementation and experiments demonstrate that the architecture has significant improvements and is feasible to be deployed.