AWS re:Invent 2025 - Amazon Aurora DSQL: A developer's perspective (DAT401)

AWS re:Invent 2025 - Amazon Aurora DSQL: A developer's perspective

Introduction to Amazon Aurora DSQL

• DSQL is a distributed implementation of PostgreSQL announced at re:Invent the previous year
• Key features:
  • Virtually unlimited scale
  • High availability
  • No infrastructure management
  • Ease of use

Cluster Creation and Connectivity

• DSQL clusters can be created quickly and easily through the AWS console
• Clusters support single-region or multi-region configurations
• Connecting to DSQL clusters can be done through the web-based query editor or AWS Cloud Shell

Developing with DSQL

• The presentation uses a starter kit with a CDK stack and Lambda function
• The Lambda function demonstrates connecting to DSQL and running simple queries
• DSQL provides secure connectivity through pre-signed URLs, eliminating the need to manage passwords

Transactions and Concurrency Control

• DSQL uses optimistic concurrency control to handle transactions
• Transactions are ACID (Atomic, Consistent, Isolated, Durable)
• DSQL provides strong snapshot isolation, with only one potential anomaly (write skew)
• Developers need to handle retries and serialization errors when conflicts occur

Building a Banking Application

• The presenters demonstrate building a simple banking application using DSQL
• The application includes functionality to transfer funds between accounts
• Handling concurrency is crucial, as multiple transactions attempting to update the same accounts can cause conflicts

Handling Conflicts and Retries

• The application includes a retry loop to handle serialization errors caused by concurrent updates
• Metrics are collected on the number of retries and duration of transactions to monitor performance
• Indexing is used to optimize query performance, with the "create index async" feature to build indexes online without blocking

Scaling and Pricing

• The application is scaled to 10,000 transactions per second without any additional configuration
• DSQL pricing is based on a "Distributed Processing Unit" (DPU) metric, which covers compute, storage, and I/O usage
• Pricing is pay-per-use, allowing the system to scale up and down without over-provisioning

Key Takeaways

• DSQL provides a fully managed, scalable, and highly available PostgreSQL-compatible database service
• Optimistic concurrency control simplifies transaction handling, but requires developers to handle retries and serialization errors
• DSQL offers features like online index creation and pay-per-use pricing to support high-scale, dynamic applications
• The presentation showcases DSQL's ability to handle complex, high-throughput workloads with minimal operational overhead