Databases Tools

The useful way to read this category is by data model, not by brand. Relational databases give you schemas, joins and transactional guarantees, and remain the correct default for most applications. Document stores relax the schema in exchange for flexibility, which helps when the shape of the data genuinely varies and hurts when it didn't need to. Key-value stores are built for speed and simple access patterns. Search engines exist because querying text well is a different problem from querying rows.

Most applications need one primary store and, eventually, a cache. The common mistake is reaching for a second database to solve a problem the first one already handles — modern relational engines do JSON, full-text search and queuing perfectly well at small and medium scale, and one system you understand deeply beats three you half-run.

The other decision is managed versus self-hosted. Managed platforms handle backups, upgrades and failover, which are exactly the things that hurt most when neglected. Self-hosting is cheaper in cash and more expensive in attention.