The paper presents a visionary research agenda for interactive exploration of extremely large databases, especially in the area of scientific data management where ingestion of multiple Terabytes of data on a daily basis is foreseen. A major innovation is required to aid scientists in their data exploration journey, where hitherto prevalent assumptions of complete and correct results should be traded against instant response and insight into the database collection at hand.
Several research directions are carved out to realize this vision. They range from engineering a novel database kernel where speed rather than completeness is the first class citizen, up to refusing to process a costly query in the first place, but providing advice on how to reformulate it instead, or even providing alternatives the system believes might be relevant for the exploration patterns observed.
The work is part of the SciLens project in the CWI research group Database Architectures (INS1).