Relational databases store data as relations (tables) with rigorously defined schemas, and they use keys, constraints, and transactions to keep data correct under concurrency. They matter because most business systems depend on predictable integrity guarantees (e.g., referential integrity, uniqueness) while still supporting high query flexibility via SQL. A useful mental model is “declare invariants + write set-based queries”: let constraints enforce correctness, and let the optimizer choose efficient execution plans. When performance surprises happen, the fastest path to clarity is usually to inspect the plan (EXPLAIN) and verify the physical design (indexes, statistics, partitioning).