dynamodb-single-table-design

DynamoDB Single-Table Design Architecture Guide

Comprehensive reference for mastering Amazon DynamoDB Single-Table Design, access pattern modeling, secondary index overloading, sparse indexes, and high-concurrency transactional operations.

---

1. Principles of Single-Table Design

Why Single-Table Design?

In DynamoDB, microsecond-latency scale is achieved by avoiding runtime joins. Instead of creating separate tables for each entity type, a Single-Table Design pre-joins heterogenous entities into a single physical table using generic Partition Keys (PK) and Sort Keys (SK).

+-------------------+--------------------+------------------------+------------------+
| PK                | SK                 | Type                   | Data Payload     |
+-------------------+--------------------+------------------------+------------------+
| USER#usr_1001     | METADATA           | UserProfile            | { name, email }  |
| USER#usr_1001     | ORDER#2026-07-01   | Order                  | { total: 49.99 } |
| USER#usr_1001     | ORDER#2026-07-18   | Order                  | { total: 89.00 } |
| ORDER#ord_9901    | METADATA           | Order                  | { status: SHIPPED|
+-------------------+--------------------+------------------------+------------------+

Access Pattern Driven Design Workflow

1. List all access patterns (Reads, Writes, Aggregations) before defining keys.
2. Design Primary Partition Keys (PK) to group related items into logical item collections.
3. Design Primary Sort Keys (SK) to enable range filtering (begins_with, BETWEEN).
4. Add Global Secondary Indexes (GSIs) for alternate access vectors.

---

2. Primary Key & Composite Sort Key Conventions

Common Key Formatting Conventions

• User Entity: PK: USER#<userId>, SK: METADATA
• User Orders: PK: USER#<userId>, SK: ORDER#<timestamp>#<orderId>
• Order Details: PK: ORDER#<orderId>, SK: METADATA
• Order Line Items: PK: ORDER#<orderId>, SK: ITEM#<lineItemId>