Database-as-a-Service for Long-Tail Science

摘要

Database technology remains underused in science, especially in the long tail - the small labs and individual researchers that collectively produce the majority of scientific output. These researchers increasingly require iterative, ad hoc analysis over ad hoc databases but cannot individually invest in the computational and intellectual infrastructure required for state-of-the-art solutions. We describe a new "delivery vector" for database technology called SQL-Share that emphasizes ad hoc integration, query, sharing, and visualization over pre-defined schemas. To empower non-experts to write complex queries, we synthesize example queries from the data itself and explore limited English hints to augment the process. We integrate collaborative visualization via a web-based service called VizDeck that uses automated visualization techniques with a card game metaphor to allow creation of interactive visual dashboards in seconds with zero programming. We present data on the initial uptake and usage of the system and report preliminary results testingout new features with the datasets collected during the initial pilot deployment. We conclude that the SQLShare system and associated services have the potential to increase uptake of relational database technology in the long tail of science.