DDIA Chapter 7: Transactions Explained | ACID, Isolation Levels & Serializability

In this video, we cover Chapter 7 of Designing Data-Intensive Applications (DDIA) by Martin Kleppmann - one of the most important chapters in the book.

We go deep into transactions: what they are, what guarantees they actually provide, and what they silently don't protect you from. This is the gap most engineers miss — and it's exactly where production bugs hide.

Here's what we cover:

► What is a transaction and why it exists
► ACID properties — Atomicity, Consistency, Isolation, Durability — explained with real examples
► Why Isolation is a spectrum, not a binary switch
► Isolation levels: Read Committed, Snapshot Isolation (MVCC), Serializable
► Concurrency anomalies: Dirty Read, Dirty Write, Read Skew, Lost Update, Write Skew, Phantom Reads
► Why snapshot isolation doesn't protect you from write skew
► Phantoms and how to handle them: predicate locks, index-range locks, materializing conflicts
► Three ways to implement Serializability: actual serial execution, Two-Phase Locking (2PL), and Serializable Snapshot Isolation (SSI)
► Stored procedures — pros, cons, and when they make sense
► Shared locks vs exclusive locks, lock upgrades, and deadlocks
► SSI: optimistic concurrency control, detecting stale reads, detecting writes that affect prior reads
► How to pick the right isolation level for your workload

By the end of this video, you'll know exactly what isolation level to use, what locks to reach for, and — most importantly — the why behind those decisions.

Link to the artifact used in the video - https://claude.ai/public/artifacts/0d065357-b342-4d28-85c1-0899dc7fa530

Видео DDIA Chapter 7: Transactions Explained | ACID, Isolation Levels & Serializability канала Nishi Ajmera