Introduction to Amazon DynamoDB

Using Data Stores in AWS

Dunieski Otano

AWS Solutions Architect

The hot partition problem

• Imagine a restaurant with one busy cashier and three idle ones
• That's what happens with poor partition key design
• Hot partition = one partition receives disproportionate traffic while others sit idle.
• Solution: High-cardinality partition keys: attributes with many unique values that distribute data evenly

What makes a good partition key

• High cardinality: Many unique values
  • userId, orderId, sessionId - Yes
  • country, status, category - No
• Even distribution: Balanced access patterns
• Predictable: Known at query time

DynamoDB table structure

• Partition key (required): Determines data distribution
• Sort key (optional): Enables range queries (e.g. userId + orderDate)
• Attributes: Additional data fields
• Primary key = Partition key OR Partition key + Sort key

Creating your first table

• AWS Console: Quick setup for testing
• boto3 (Python): Programmatic table creation
• Key decisions: Partition key, sort key, billing mode

Composite keys and sort key queries

• Composite key: Partition key + Sort key
• Range queries: Query within date ranges, numeric ranges
• Operators: =, <, >, <=, >=, BETWEEN, begins_with
• Example: Get orders for user in last 30 days > $100

Capacity modes: on-demand vs provisioned

• On-Demand: Pay per request, auto-scales
  • Best for: Unpredictable traffic, new applications
• Provisioned: Set RCU/WCU, lower cost at scale
  • Best for: Predictable traffic, cost optimization
• Can switch modes once per 24 hours

Let's practice!

Using Data Stores in AWS