Courses / Course 01 · Databases
You end up withA durable key-value store you wrote, with a B+tree, crash recovery and transactions, a small SQL layer on top, and enough knowledge of PostgreSQL's internals to read its source.
Why build one when an agent can write it for you? Because the next time your agent says “just add an index” or “wrap it in a transaction”, you’ll know what it’s really asking the disk to do. Hacker News has voted these tutorials up again and again for over a decade. Every step below shows its HN score and quotes what HN said.
How to read the receipts.
An animated walk from “append to a file” to indexes to log-structured merge trees. Do it first: it gives you the vocabulary for everything after.
“It’s a real test of how much you know to start anything from scratch.”
One long article covering the whole relational engine: query parser, optimizer, joins, buffer pool, locks, logging. Read it once, now, without trying to remember everything. You’ll come back to it.
“As complicated as they look from the article, it is just scratching the surface… there is probably more hardcore computer science per line of code than any other software system of similar scope.”
The simplest durable database worth building: an append-only log on disk with an index in memory. The project comes with tests that guide each stage.
“In short, keys are stored in a hash table in RAM, k/v pairs are written to log files. Dead simple, but powerful enough.”
“That whole experiment sent me on a different career trajectory.”
The most-resubmitted database tutorial on HN. You build a REPL, a pager, then an on-disk B-tree with node splitting.
“Writing a database is a good exercise I could recommend to everybody.”
Editor’s note: the series ends at Part 14, “Splitting Internal Nodes”. It has no deletes, no transactions and no crash recovery, and the author marks it “no longer under active development”. Part 15 points readers to CodeCrafters through a referral link. Step 6 picks up where it stops: deletes, crash recovery and the free list are all in the free part of that book. (Checked 2026-11-08.)
Free · ~1 evening each. Read them with your step-4 code open.
“I really appreciate how Julia Evans enjoys experimenting in programming… compared to so many practitioners who just read the books and often uncritically ‘believe’ the written texts.”
Also recommended in the Fly.io thread: SQLite’s own design documents, especially “Atomic Commit In SQLite” and “How To Corrupt An SQLite Database File”.
“I recommend the excellent articles by the SQLite authors themselves … They go into the details of challenges SQLite faces due to how hardware and OSs work or not work and how they solve them.”
The core of the track. It comes in two parts:
The whole book comes to about 3,000 lines with no dependencies, according to the 2025 edition’s announcement (422 points on its own).
Free route to the endpoint: do Part I. Step 7 then gives you the SQL layer, and the concurrency-control part of step 11 covers transactions. Part II is the paid shortcut that does both in one codebase. (Terms checked 2026-11-08.)
“I also recommend this free online book to build a database.”
Alternatives HN suggested in these threads:
A lexer, a parser, and an executor for CREATE, INSERT and SELECT … WHERE. This is the query-evaluation half that the storage tutorials skip.
“Probably the first article Ive ever read about lexical parsing with code that i have actually understood - and I dont even program in golang.”
This one covers networking rather than storage: sockets, an event loop, a hash table, sorted sets, TTLs.
”…Redis is the quintessential network service: By implementing it, you learn about socket handling, event loop for a specific runtime, threading models, data representation, concurrency…”
In Rust instead: Tokio’s mini-redis tutorial (https://tokio.rs/tokio/tutorial; 72 points, submitted 4×). One warning from its thread: “Rust async programming is a bad way to learn about Rust, and also is a bad way to learn about async programming.” (sanxiyn, 2020 · comment)
One query traced through Postgres’s parser, analyzer, rewriter, planner and executor, all in the C source. Read it before step 9 so the big book has a skeleton to hang on.
“He does a great job of combining high level explanations, visual aids, and gritty details.”
The highest-scoring database resource in the dataset. It covers how a real engine handles the things your toy skipped: heap pages, MVCC, vacuum, the planner, the WAL, replication.
“This is the thing I love about PostgreSQL, you can find out everything about the database.”
Follows a single transaction through the Postgres source, with diagrams.
“This is precisely the kind of article I hope for when I browse the front page… It is clearly the product of a lot of effort and craft.”
The university course that ties steps 3–10 together. It has projects (buffer pool, B+tree index, query execution) on the BusTub teaching database.
“Andy Pavlo’s lectures are one of the best database internals resources available for free on the web.”
“These kinds of posts are my favorite part of HN… miles above the basic posts around the internet you get with searching – but not quite a full book yet.”
Next steps HN recommended in the same thread:
| Book | HN signal | Cost | Read it during |
|---|---|---|---|
| Designing Data-Intensive Applications · Martin Kleppmann | 610 points over 6 submissions; named in 680 comments | Paid book | Stages 3–5 |
| Database Internals · Alex Petrov | 308 points over 2 submissions; named in 343 comments | Paid book | Stages 3–4 |
| Readings in Database Systems (“Red Book”) · Bailis, Hellerstein, Stonebraker · http://www.redbook.io/ | 857 points over 14 submissions (2015–2025) | Free | After Stage 4 |
| Architecture of a Database System · Hellerstein, Stonebraker, Hamilton (2007) · https://dsf.berkeley.edu/papers/fntdb07-architecture.pdf | 241 points over 14 submissions | Free paper | Stage 4 |
“It will not appeal to the absolute novice to be sure. But for anyone else who has worked on systems for moving data… this book will show you how things (probably stuff you’ve done bits of pieces of) fit together…”
“The depth of a text book with the approachability of a typical working programmer’s book.”
“I definitely wouldn’t approach this like I would approach a textbook: it’s definitely not meant to be friendly introductory material. That said, once you have a bit of background, it’s a goldmine of a survey.”
“I found this paper ‘Architecture of a Database System’ [1] to be an excellent starting point…”
Build one to learn. Don’t put one into production unless the database is the project.
“…any software project that includes writing a database is, de facto, a database project.”
Checked against HN on 2026-11-08 (backlog_receipts.json). Not promoted:
Promoted: the Red Book and “Architecture of a Database System” (theory shelf), Shaughnessy (Stage 4 warm-up), Mini-LSM (step 6 alternative) and Tokio mini-redis (step 8 alternative).
tracks/databases/receipts.json (all resources cited above) and backlog_receipts.json (the backlog check). Scripts are in phase1-stories/scripts/.