Data

Why databases exist — programs need data to survive restarts, and multiple programs need to share the same data without corruption.

The relational model — tables, rows, columns, keys. SQL as the language to ask questions. Why this 50-year-old idea still dominates: it separates "what you want" from "how to get it".

Transactions and ACID — the four promises a database makes (Atomicity, Consistency, Isolation, Durability). Isolation levels as how strict the isolation really is. MVCC as the trick that lets readers and writers coexist.

Indexes — without them, every query reads every row. B-trees (the workhorse), LSM trees (the write-optimized alternative). The trade-off: faster reads vs faster writes.

Beyond relational — document stores (flexible schemas), graph databases (relationships first), time-series (optimized for timestamped data), vector stores (nearest-neighbor search for embeddings). Each exists because one access pattern matters more than the rest.

Data engineering — getting data from where it's born to where it's useful. Batch (process all at once, periodically) vs stream (process as it arrives). Warehouses (structured, analytical queries) vs lakehouses (raw + structured, one system). The idea of lineage: knowing where data came from and what happened to it.