A municipal asset registry must keep device rows consistent with site and owner tables, enforce foreign keys, and support ad-hoc joins under transactions. Which database type best fits that stated need?

A relational store that supports joins, constraints, and transactions best fits a municipal asset registry that must keep device rows consistent with site and owner tables, enforce foreign keys, and support ad-hoc joins under transactions. Relational engines are built for integrity across related entities: foreign keys prevent orphan device rows, transactions keep multi-table updates atomic, and SQL joins answer ownership and site questions without application-level glue. When the stated requirement is ledger-like consistency across tables, relational is the matching official database type.

A time-series store optimized for append-only counter samples answers heartbeat and windowed metric questions, not foreign-key integrity across device, site, and owner. A graph database chosen only because the network has a topology diagram optimizes for path traversal; it is not justified merely by a picture if the workload is constrained joins and transactions. A document store with no integrity constraints across collections can hold flexible JSON but drops the cross-collection guarantees the registry explicitly requires.

Exam caveat: relational fit assumes you actually need constraints and joins—do not force every inventory blob into rigid tables when the access pattern is whole-document cache. Operational check: attempt an insert of a device row with a non-existent site foreign key in a staging registry and confirm the engine rejects it; then run a transactional update that moves ownership and verify a mid-failure rollback leaves no half-applied rows.