CONIX Publication

Programming distributed applications in the IoT-edge environment is a cumbersome challenge. Developers are expected to seamlessly handle issues in dynamic reconfiguration, routing, state management, fault tolerance, and heterogeneous device capabilities. We introduce DDFlow, a macroprogramming abstraction and accompanying runtime that provides an efficient means to program high-quality distributed applications that span a diverse and dynamic IoT network. We describe the programming model and primitives used to isolate application semantics from arbitrary deployment environments. Using DDFlow leads to portable, visualizable, and intuitive applications. The accompanying system runtime enables dynamic scaling and adaptation, leading to improved end-to-end latency while preserving application behavior despite device failures.