Leveraging Dataflow as an Intermediate Representation for Portable Edge Deployments

In this paper, we investigate how to leverage the dataflow programming model to facilitate the deployment of multi-scale (sensors-edge-cloud) IoT applications end-to-end. Today, IoT deployment management is error-prone, tedious, and manual - yet deployment configuration can have a significant effect on application energy consumption and performance. Our approach, called deployment-as-code, provides programming directives that distribute application components across heterogeneous devices (including microcontrollers), aid management of the application lifecycle, and facilitate optimization in ways not available from traditional programming systems. We implement our approach by extending an open-source dataflow programming system for IoT, which we use to evaluate the energy consumption of different deployment configurations and IoT applications.