DynamoDB
Amazon DynamoDB from the inside — the data model and partitioning, capacity and throughput math, replication and consistency, the request path, lock-free ACID transactions, the storage engine, and what it kept and changed from the 2007 Dynamo paper.
Module 0 — Glossary & Mental Models
The core DynamoDB vocabulary and mental models — tables, items, partition and sort keys, secondary indexes, capacity units, and replication.
Module 0b — Under the Hood
How DynamoDB physically stores and serves data — partitions as replication groups, WAL and B-tree storage nodes, and the MemDS metadata plane.
Module 1 — Data Model & Partitioning Fundamentals
How DynamoDB models and partitions data — simple vs composite keys, attribute types, the 400 KB item limit, and hashing keys onto partitions.
Module 2 — Capacity, Throughput & Warm Throughput
On-demand vs provisioned modes, RCU/WCU math with per-item rounding, partition limits, throttling, and warm throughput.
Module 3 — Replication, Quorums & Consistency
How DynamoDB replicates each partition across three AZs with single-leader Multi-Paxos, 2-of-3 write quorums, and consistency modes.
Module 4 — Partitioning and Request Path
How DynamoDB partitions data with hash-range partitioning and routes every request through stateless routers and the MemDS metadata service.
Module 9b — How DynamoDB Maintains ACID Transactions
Serializable ACID transactions without locks — optimistic concurrency control, timestamp ordering, and a two-phase prepare/commit protocol.
Module 10 — Storage Engine Internals
Inside a DynamoDB storage node — the write-ahead log, B-Tree, 2-of-3 quorum replication, partition splits, GSIs, and Multi-Paxos leader election.
Module 11 — Dynamo Heritage and What DynamoDB Changed
What DynamoDB inherited from the 2007 Dynamo paper and what it redesigned — from leaderless vector clocks to single-leader write ordering.