DBMS

How databases store data on disk, how indexes work internally, ACID properties and what each guarantees, transaction isolation levels and their anomalies, normalization and when to denormalize, how the query engine parses and executes SQL, replication models for scaling reads, and the CAP theorem trade-offs behind NoSQL systems.

Heap files, pages, B-tree and LSM-tree, clustered vs non-clustered index, atomicity, consistency, isolation, durability, WAL, MVCC, dirty read, non-repeatable read, phantom read, isolation levels, 1NF/2NF/3NF/BCNF, query plan, cost-based optimizer, primary replica, read replica, sharding, CAP theorem, eventual consistency.

SQL tells the database what to do; DBMS internals explain how it does it — and why it sometimes does it slowly or incorrectly under concurrent load. Engineers who understand transactions and isolation levels write code that doesn't corrupt data under race conditions. Engineers who understand storage and indexes make choices that stay fast at scale.

System design interviews probe DBMS knowledge directly — "how would you scale this database?", "what isolation level do you need here?", "why is this query slow?". Senior and staff roles expect you to reason about consistency trade-offs, choose between replication strategies, and explain what ACID actually guarantees at each isolation level.