A Domain-Driven, Generative Data Model for Big Pet Store

Generating large amounts of semantically-rich data for testing big data workflows is paramount for scalable performance benchmarking and quality assurance in modern machine-learning and analytics workloads. The most obvious use case for such a generative algorithm is in conjunction with a big data application blueprint, which can be used by developers (to test their emerging big data solutions) as well as end users (as a starting point for validating infrastructure installations, building novel applications, and learning analytics methods). We present a new domain-driven, generative data model for Big Pet Store, a big data application blueprint for the Hadoop ecosystem included in the Apache Big Top distribution. We describe the model and demonstrate its ability to generate semantically-rich data at variable scale ranging from a single machine to a large cluster. We validate the model by using the generated data to answer questions about customer locations and purchasing habits for a fictional targeted advertising campaign, a common business use case.