The project “KoDa-EFC“ is a micro-project as part of the Software Campus Program supported by the BMBF.
The project places complex event processing operators onto the edge-fog-cloud (EFC) continuum. Response latency and cost must be minimized.
Complex event processing (CEP) processes streams of data tuples. It derives complex, higher-order information from primitive events.
A network of operators, the operator graph, thereby stepwise processes the incoming tuple streams. It aggregates tuples and searches for patterns in the streams.
With the realm of fog and edge processing, more computation power is available closer to the sources and consumers of the streams. Examples are IoT devices, on-premise infrastructure, smartphones.
Yet, huge processing capacity and flexibility still keep the established cloud computing interesting for CEP. Hence, the project develops techniques to execute distributed CEP applications along the EFC continuum. It thereby faces the challenges of heterogeneous hardware, especially in the edge and fog layers, and dynamics in processing performance and workload typical for long-running stream processing applications.
In the first phase, we focus on building a framework legible to execute CEP applications on a wide range of processing nodes. The framework includes load managers that distribute the incoming workload following the current processing performance.
In the second phase, we optimize the load management that ensures well-balanced execution. It handles the dynamics in the EFC infrastructure, e.g., due to battery drain, overload, or node movement. We aim at a distributed solution to avoid long-distance control messages sent to a centralized control instance. We investigate different operator types and load distributions as well as different performance dynamics.