How Databases Work - For Developers, Accidental DBA's and Managers
summary
Postgres internals explained as concrete mechanics not marketing: data lives in 8KB blocks with row pointers, parsing → planner → executor produces plans like nested‑loop joins that can turn an index hit into repeated full scans, concurrency via MVCC so readers don’t block writers and vice versa, durability via sequential write‑ahead log (WAL) flushed at commit and replayed on recovery, checkpoints truncate WAL (configurable e g 5–10 minutes), autovacuum reclaims dead row versions and BGWriter prewrites dirty buffers.
AI-written summary · about
chapters (9)
- 0:00Start
- 4:43ACID
- 8:49Application Interaction
- 11:35Backend Operations - TokenizelParse
- 14:24Backend Operations - Parse Analysis
- 15:16Backend Operations - Plan Opimize
- 20:49Backend Operations - Execute
- 24:07ConcurrentCooperating Operations
- 34:24Full Monty
related talks
48mA Deep Dive Into Database Internals
47mWhat If We Opened PostgreSQL's Hood?
11mThe Seven Principles for Making a PostgreSQL DB Work Fast
21mA Guided Tour Through Postgres Internals with Elizabeth Garrett Christensen
39mDeep dive into PostgreSQL internal statistics
51mHow does PostgreSQL actually work
29mPostgres Demystified
52mUnderstanding PostgreSQL Query Plans (aka EXPLAIN)