AWS re:Invent 2025 - Deep dive into Amazon Aurora and its innovations (DAT441)

Summary of AWS re:Invent 2025 - Deep dive into Amazon Aurora and its innovations (DAT441)

Overview of Amazon Aurora

• Amazon Aurora is a cloud-native relational database that combines the speed and availability of high-end commercial databases with the simplicity and cost-effectiveness of open-source.
• It is fully managed, fully compatible with MySQL and PostgreSQL, and provides tools for integrating with serverless machine learning applications.
• The most recent addition to the Aurora family is the Aurora DSQL (Distributed SQL) engine.

Aurora Architecture

• Aurora storage is the key component that makes Aurora special, providing three-way replication across three availability zones out of the box.
• The storage nodes write database log records, which are then turned into database pages, eliminating the need for traditional database operations like checkpoints, full page writes, and log archival.
• Aurora provides automatic failover capabilities, where a replica can be automatically promoted to the primary node if the original primary node fails.
• Aurora supports multi-region deployments through global database, where the storage is asynchronously replicated between regions, and read-only replicas can be created in the secondary regions.
• Aurora provides a global endpoint that automatically routes traffic to the current primary region, and supports fast failover and switchover operations between regions.

Local and Global Write Forwarding

• Aurora supports local write forwarding, where writes from a secondary availability zone are forwarded to the primary node for execution.
• Aurora also supports global write forwarding, where writes from a secondary region are forwarded to the primary region for execution.
• These features come with consistency considerations, as the application may need to handle asynchronous replication and the resulting consistency trade-offs.

Aurora Storage Internals

• Aurora's storage nodes handle the conversion of log records into database pages, enabling features like continuous backup to S3 and peer-to-peer repair of missing log records.
• Aurora introduced an "IO-optimized" storage configuration, which provides a more predictable pricing model and improved performance, especially for latency-sensitive workloads.
• The IO-optimized configuration includes features like a tiered cache, which uses local NVMe storage to cache frequently accessed data and improve read performance.

Aurora Postgres and MySQL Innovations

• Both Aurora Postgres and Aurora MySQL track the latest upstream versions, with support for new features and performance improvements.
• Key innovations include dynamic data masking, global database secondary readers, and the new "create with express configuration" feature that allows creating Aurora clusters in seconds.

Integrating Aurora with Serverless and AI/ML

• Aurora's compatibility with MySQL and Postgres allows it to be easily integrated with serverless applications and AI/ML frameworks like Hugging Face and Anthropic.
• The presentation showcased how Aurora can be used as the memory store for large language models, taking advantage of its scalability and low-latency access.
• Aurora also provides Model Context Protocol (MCP) servers that allow natural language queries to be translated into SQL, making it easier for developers to interact with the database.

Aurora DSQL

• Aurora DSQL is the newest addition to the Aurora family, providing a distributed SQL engine with an active-active architecture, as opposed to the active-passive model of Aurora Postgres.
• DSQL uses optimistic concurrency control to handle concurrent writes, and its distributed block store allows for independent scaling of the read and write paths.
• The presentation compared the key differences between Aurora Postgres and Aurora DSQL, highlighting the trade-offs between the active-passive and active-active models.

Conclusion

The presentation provided a comprehensive overview of the latest innovations in Amazon Aurora, covering the architecture, storage internals, feature updates, and integration with serverless and AI/ML use cases. The depth of technical details and real-world examples showcased Aurora's capabilities in addressing the evolving needs of modern data-driven applications.