Adaptive Big Data Pipeline
Descripción
Over the past three decades, data has exponentially evolved from being a simple software by-product to one of the most important companies’ assets used to understand their customers and foresee trends. Deep learning has demonstrated that big volumes of clean data generally provide more flexibility and accuracy when modeling a phenomenon. However, handling ever-increasing data volumes entail new challenges: the lack of expertise to select the appropriate big data tools for the processing pipelines, as well as the speed at which engineers can take such pipelines into production reliably, leveraging the cloud. We introduce a system called Adaptive Big Data Pipelines: a platform to automate data pipelines creation. It provides an interface to capture the data sources, transformations, destinations and execution schedule. The system builds up the cloud infrastructure, schedules and fine-tunes the transformations, and creates the data lineage graph. This system has been tested on data sets of 50 gigabytes, processing them in just a few minutes without user intervention.ITESO, A. C.