A guided tour of a real database engine
On the surface, db.Insert("Users", row) looks trivial. Under the hood, a stack of specialized layers cooperate in milliseconds to make that row findable, durable, and safe from crashes. This course traces a single request all the way down to the bytes on disk — using a real, working C#/.NET engine you can read end to end.
You write apps on top of databases every day. This is the layer you never had to think about — until a query went slow, data vanished after a crash, or an AI suggested something that smelled wrong.
Every concept is anchored to real code from a working engine, a plain-English translation beside it, and an interactive widget you can poke. No hand-waving.
When you can name the layers — storage, index, transaction, query — you can debug production and push back on AI with specifics instead of guesses.
The curriculum
Work top to bottom — each module zooms into one layer of a real database, building on the last. Roughly 12–18 minutes each.
Trace one INSERT through all five layers. Learn why a database is a stack of specialists, each owning one concern.
The storage floor plan: fixed 4 KB pages, 8-page extents, and the LRU cache that keeps hot pages in RAM.
Find one row among millions in ~23 comparisons. Build a live tree, watch nodes split, and run range scans.
Why the same query can be instant or glacial. Run a live planner and see when the index is used vs. a full scan.
How databases never lose your money. The write-ahead log, ACID, and crash recovery — log first, apply after.
Many hands, one ledger. Reader-writer locks, lock ordering, and how to avoid the deadly embrace of deadlock.
How real databases let readers never block writers — row versioning, snapshots, and the isolation levels that decide what you see.
You can't open the hood mid-flight — you read the gauges. Drive a live metrics dashboard and watch structured logs stream past.