Amazon DynamoDB - A Scalable, Predictably Performant, and Fully Managed NoSQL Database Service

Explore a 20-minute conference talk from USENIX ATC '22 detailing Amazon DynamoDB, a scalable, predictably performant, and fully managed NoSQL database service. Dive into the architecture, evolution, and operational experiences of this cloud database relied upon by hundreds of thousands of customers. Learn about DynamoDB's fundamental properties, including consistent performance, availability, durability, and serverless experience. Discover how the system handled massive-scale operations during Amazon Prime Day, peaking at 89.2 million requests per second. Gain insights into the challenges faced and solutions implemented for fairness, traffic imbalance, monitoring, and automated system operations. Understand the importance of reliability in database services and how DynamoDB's design continues to evolve to meet increasing customer demands. The talk covers key topics such as partitioning, admission control, overprovisioning, uniform distribution, and lessons learned from operating at scale.

Introduction
Fully Managed NoSQL
DynamoDB Stats
Dynamo vs DynamoDB
DynamoDB Architecture
Why DynamoDB
Partitioning
Admission Control
Overprovisioning
Uniform Distribution
Global Admission Control
Partition Burst
Lessons Learned
Conclusion