The domain of Visual Analytics has emerged with a charter to support interactive exploration and analysis of large volumes of (often dynamic) data. A common feature shared by all the visual analytics applications developed so far is the reliance on ad-hoc and custom-built mechanisms to manage data: they re-implement their own in-memory databases to support real-time display and interactive feedback and analytical algorithms (e. g., clustering, multidimensional projections, specific data analyses) to overcome the delay required to exchange data with specialized analytical environments, such as Matlab, R, and the myriad of more specialized systems and command-line programs. This article presents visual analytics scenarios requiring a new generation of databases to support and foster the new era of interactive analytical environments. This new generation would relieve visualization researchers from sub-optimally re-implementing database technologies. We argue that the new services required to support interactive explorations present research challenges to the database community and can increase the utility and simplicity of integration of the new generation of databases for data-intensive applications.