Adaptive Computation Offloading for Heterogeneous IoT Applications: An Experimental Study.

Internet of Things (IoT) applications are heterogeneous in terms of the deployed hardware, developed protocols, and requirements of each of these applications. The decision whether to process such applications at the sensor nodes or at a higher layer (i.e., Edge, Fog, or Cloud) heavily depends on those heterogeneity features. In this paper, we propose a number of adaptive schemes to address the local processing vs. offloading of IoT applications based on their energy consumption and execution time. To this end, we deploy six IoT applications on real sensor nodes, measure their energy consumption and execution time, and then decide where to process the application based on the measurements. We study the performance of the proposed schemes in terms of energy efficiency and execution time. We also compare their performances to selected non-adaptive schemes.