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. To-

day, 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 program-

ming directives that distribute application components across

heterogeneous devices (including microcontrollers), aid manage-

ment 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.