Many applications depend on diverse data structures that must be persisted and queried. Developers face a dilemma: they can cram different kinds of data – such as relational, graph-structured, and semi-structured – into a single store, or they can incur the operational cost of provisioning and running multiple DBMSs and their associated technology stacks.

This talk will explore a new approach for overcoming this dilemma: using a transactional ordered key-value store where multiple logical data models can be mapped to the same storage substrate. This approach provides the flexibility of polyglot persistence without the latter's operational overhead. A database that effectively supports multiple data models can provide a better option than more specialized solutions such as graph or document-based storage formats. In particular, along with the operational advantages, multiple query/traversal patterns can be supported with the same physical representation.