Subsequent projects

Prof. Dr. Markus U. Mock
University of applied sciences Landshut
Fakultät Informatik

Prof. Chandra Krintz
University of California, Santa Barbara
Department of Computer Science

Serverless Computing for IoT: from Edge to Cloud and Back

The serverless computing paradigm has quickly gained much interest as a flexible and scalable way of performing storage and compute functions in the cloud without complicated deployment or management of computing infrastructure. Due to bandwidth and latency requirements, bringing this paradigm to the rapidly growing Internet of Things (IoT) does not work out of the box. At the same time, edge computing per se is not a solution as it sets IoT deployments back to managing their computing infrastructure. In this project, we want to explore how to bring serverless computing to IoT deployments that are bandwidth-constrained and latency-sensitive, such as smart farm environments, so that functions as a service become available from the edge to the cloud as seamlessly as possible. To achieve this, we are investigating what language and systems primitives are necessary to build robust IoT data collection and analytics pipelines that can function without the need for operating complex compute infrastructure.


Primary project: Data Analytics of Soil Health Data


Final Report

The project's original idea was to explore how to bring serverless computing to IoT deployments so that functions as a service become available from the edge to the cloud as seamlessly as possible. This was supposed to be done by investigating language and system primitives necessary to build a robust IoT data collection and analytics pipeline.

Unfortunately, due to the pandemic's onset, it was impossible to progress on the initially proposed goal. The California-wide shutdown made all intense collaborative activities impossible as UCSB professors were involved full-time in keeping instruction working at a minimum level, and since the run-time system was supposed to come from UCSB, we had to pivot the project from investigating language and run-time mechanisms for a robust IoT platform to instead leverage the versatility of the cloud-based serverless approach in a different application domain, namely energy-management systems.

We were able to achieve two main results. First, we demonstrated the versatility of a serverless architecture for data acquisition, storage, and visualization by creating a fully functional pilot system at the HAW Landshut, with almost one-hundred measurements, e.g., real-time electricity consumption in various buildings, real-time water consumption, real-time weather data and size of electricity production of the university’s photovoltaic installation. This reference architecture was published in the peer-reviewed European Conference of Service-Oriented and Cloud Computing conference as Upilio: Leveraging the Serverless Paradigm for Building a Versatile IoT Application by Markus Mock and Stefan Arlt, 9th European Conference on Service-Oriented and Cloud Computing, 22 - 24 March 2022, Wittenberg, Germany, (run virtually).

As a second main result, we demonstrated that everyday data analytics operations (OLAP), including on-demand visualization, could be implemented efficiently and cost-effectively using the serverless paradigm. In addition to manual analyses, we were also able to demonstrate the efficient implementation of simple computational intelligence in the system, for instance, the automatic detection of solar electricity production anomalies, e.g., due to malfunctioning inverters. This was achieved by “synthesizing” a sensor value from the normalized global radiation of the local weather station combined with the normalized sensor of PV power. How to realize such real-time computational intelligence serverlessly was presented at the (peer-reviewed) IEEE Computational Intelligence Conference in Toci: Computational Intelligence in an Energy Management System by Florian Huber and Markus Mock, 2020 IEEE Symposium Series on Computational Intelligence, December 1-4, 2020, Canberra, Australia (run virtually).


Privacy Notice