Graphical Spark Programming in IoT Mashup Tools
Tanmaya Mahapatra, Ilias Gerostathopoulos, Christian Prehofer, Shilpa Ghanashyam Gore · 2018
With the unprecedented rise in the number of IoT devices, the amount of data generated from sensors is huge and often demands an in-depth analysis to acquire suitable insights. Mashup tools, used primarily for intuitive graphical programming of IoT applications, can help both for efficiently prototyping and also data analytics pipelines. In this study, we focus on the tight integration of data analytics capabilities of Spark in IoT mashup tools. The main challenge in this direction is the presence of a wide range of data interfaces and APIs in the Spark ecosystem. In this study, we contribute to current applications by (i) providing a thorough analysis of the Spark ecosystem and selecting suitable data interfaces for use in a graphical flow-based programming paradigm, (ii) devising a novel, generic approach for programming Spark from graphical flows that comprises early-stage validation and code generation of Java Spark programs. The approach is implemented in aFlux, our JVM-based mashup tool and is evaluated in three use cases showcasing the machine learning and stream analytics capabilities of Spark.