In this talk, I will make the case for an adaptive and declarative data management infrastructure supporting such formats. I will first describe TrajStore, a storage system for very large trajectory datasets. I will outline the general back-end architecture of TrajStore, and how the system adaptively co-locates trajectory subsegments to facilitate query resolution. I will then describe SciDB, a distributed data management system for data-intensive scientific applications.
Finally, I will outline a generic architecture to handle non-relational datasets in a declarative fashion.