Engineering big data solutions

Structured and unstructured data in operational support tools have long been prevalent in software engineering. Similar data is now becoming widely available in other domains. Software systems that utilize such operational data (OD) to help with software design and maintenance activities are increasingly being built despite the difficulties of drawing valid conclusions from disparate and low-quality data and the continuing evolution of operational support tools. This paper proposes systematizing approaches to the engineering of OD-based systems. To prioritize and structure research areas we consider historic developments, such as big data hype; synthesize defining features of OD, such as confounded measures and unobserved context; and discuss emerging new applications, such as diverse and large OD collections and extremely short development intervals. To sustain the credibility of OD-based systems more research will be needed to investigate effective existing approaches and to synthesize novel, OD-specific engineering principles.