Investigating a Streaming Analytics Framework for Data Analytics Applications in the Aircraft Cabin

Abstract

Aircraft cabin operations shift towards data-driven processes. Cabin-wide multi-system communication networks are introduced to share required data for corresponding novel data-driven applications. Examples are data-driven predictive maintenance applications to reduce the downtime of systems and increase the period of scheduled maintenance or video analytics usage to detect a strained or unruly atmosphere amongst passengers. These applications require a network for transport of associated data and resources for actual computation. Costs and weight have always been the most important factors deciding if new services are introduced within the aircraft cabin. Thus, re-using hardware with free computation capacity that is already installed in the aircraft cabin can target both aspects, weight and costs. Examples for such hardware resources could be the In-flight Entertainment (IFE) equipment being installed in every seat. By means of distributed computing theses resources can be combined in order to solve a computational task. In this paper, the Apache Spark Streaming Framework is investigated as an example for a distributed computing framework with respect to its suitability for being deployed in the aircraft cabin. For this purpose, it is evaluated how Spark scales with resource-constrained computing nodes. As this type of framework is designed for a robust execution of tasks, it can also be used for non-safety, but business-critical applications. In addition, Spark is a well-established commercial off-the-shelf (COTS) software, thus, possible advantages of such software like a reduction of non-recurrent engineering costs and support for developers creating applications for the aircraft cabin are also considered.