Title

AWS re:Invent 2023 - Schema design for fast applications (DAT347)

Summary

The speaker has extensive experience in NoSQL and worked with AWS DynamoDB and Amazon retail projects.
The focus of the talk is on designing schemas for NoSQL databases, particularly MongoDB, to optimize performance.
Good performance is defined by response time, throughput, efficiency, and scalability.
The speaker emphasizes the importance of aligning the schema with the workload and how data is accessed.
Common anti-patterns in schema design include over-embedding and over-normalization.
The speaker discusses the use of patterns like the computed pattern and subset pattern to optimize data access.
Real-world examples are provided, including an order processing workflow and a real-time trading example.
The speaker concludes that poor database performance is often due to inefficient schema or query design rather than the technology itself.

NoSQL databases like MongoDB offer flexibility in schema design, allowing for optimization based on specific access patterns.
Embedding data in documents can lead to performance gains by reducing the need for joins, but it must be done judiciously to avoid performance issues from over-embedding.
The speaker highlights the importance of understanding high-frequency access patterns and optimizing for those, rather than less frequent operations.
Schema validation is crucial in NoSQL databases to ensure data integrity and avoid "Wild West" scenarios with unstructured data.
Real-world examples demonstrate that rethinking data modeling to align with NoSQL principles can lead to significant performance improvements.
The session underscores the importance of enablement and education in helping developers transition from traditional relational databases to NoSQL databases effectively.