Framing a Sustainable Architecture for Data Analytics Systems: An Exploratory Study

Abstract

Data analytics systems (DASs) with big data capabilities have started playing a promising role in online service ecosystems and large-scaled interconnected systems of many enterprises. The rapid development of analytics models and technologies, along with affordable infrastructures and accumulated data repositories, leads to encouraging expectations on DAS, while also bringing challenges in terms of how to deal with the increased development complexity. However, systematic methodologies for designing a sustainable DAS are still missing. To harness the dynamics raised by technology evolution, ambiguous requirements, under-explored data environments, and so on, framing a sustainable software architecture turns out to be a critical task. By exploring the complex nature of DAS, we propose a novel approach, sustainable architecture development for DAS (SstAD-DAS), to provide practical guidelines for architecture development. A shock absorber mechanism is presented to harness the dynamics of DAS and facilitate the development of a sustainable architecture, the “long decision chain” challenges are handled with a generic process model, and collaborations and responsibilities of participants are suggested to enable better model implementation. SstAD-DAS allows architects to accommodate the long decision chain, leverage skill sets from multiple contributors, and evaluate architectural decisions continuously. Finally, this paper demonstrates the capability and usability of SstAD-DAS by sharing experiences and observations from the continuous development of an intelligence analysis system.