Configurable and Extensible Multi-flows for Providing Analytics as a Service on the Cloud

摘要

Compared to traditional analytics deployment models, cloud-based solutions for business analytics provide numerous advantages such as reduction of a large upfront infrastructural cost and the efforts to setup an in-house analytics team. Such advantages of cloud-based service delivery make it particularly attractive for small and medium businesses. In spite of these advantages, analytics penetration has been low particularly in developing regions such as India and China due to many other factors. In this paper, we propose pre-packaged configurable workflows for analytics as a means of endearing cloud-based analytics to customers, with a special focus on small and medium businesses in developing regions. We introduce the concept of configurable multi-flows that make it easy for non-technical personnel to use and customize without being aware of the technical details of the various operators involved in the workflow. Multi-flows comprise of an overlap of multiple possible workflows and are easily extensible to include more variations to support the evolving needs of customers non-disruptively and incrementally. We detail a case-study of the Retail sector where an extensive survey of retail businesses in India revealed that configurable pre-packaged workflows may indeed help improve market penetration. We then identify common analytics needs of retail customers, and detail how such tasks can be expressed as configurable multi-flows. Further, we describe a fully functional implementation of our system that supports configurable multi-flows for analytics. Finally, we illustrate the ease-of-use of configurable multi-flows with the use of multiple screenshots.