AWS re:Invent 2025 - Amazon Aurora HA and DR design patterns for global resilience (DAT442)

Resilient Database Design with Amazon Aurora

Defining Resilience

• Resilience is the ability of a workload to recover from disruption, dynamically acquire resources to meet demand, and mitigate issues like misconfiguration.
• Resilience has two key pillars: availability and disaster recovery (DR).
• Availability is the proportion of time a workload is available for use, often measured as historical uptime.
• Disaster recovery refers to techniques used to recover a workload when something goes wrong, measured by recovery time objective (RTO) and recovery point objective (RPO).

Amazon Aurora MySQL and PostgreSQL

• Aurora MySQL (AMS) and Aurora PostgreSQL (APG) are relational database services that combine the speed and availability of commercial databases with the simplicity and cost-effectiveness of the cloud and open source.
• They are fully compatible with the open-source MySQL and PostgreSQL engines, allowing existing applications and tools to run without modification.

Handling Database Node Failures

• In a self-managed database, a single node failure can lead to significant downtime and data loss.
• Aurora addresses this with its storage layer, Grover, which replicates data across 3 Availability Zones (AZs) with 6 copies, providing durability even if an entire AZ fails.
• Aurora's log-structured storage allows for fast, zero-RPO backups and point-in-time restores without impacting performance.
• To ensure high availability, Aurora allows you to add read replicas that can be failed over to quickly in the event of a primary node failure.

Scaling Performance

• Aurora automatically scales storage capacity and performance as needed, without manual provisioning.
• You can add read replicas to offload read traffic and isolate workloads, improving performance.
• Aurora provides connection pooling and load balancing features to efficiently manage connections across the cluster.

Multi-Region Resilience with Aurora Global Database

• Aurora Global Database allows you to replicate your database asynchronously across up to 10 secondary AWS Regions.
• This provides durability and high write availability across multiple regions, with the ability to fail over the primary region if needed.
• The global endpoint and automated failover capabilities make it easy to manage the multi-region setup.
• Switchovers between primary and secondary regions can be performed with minimal downtime and data loss.

Amazon Aurora Serverless v2 (DSQL)

DSQL Architecture

• DSQL separates the compute (Query Processors) and storage components, allowing for independent scaling and high availability.
• The Adjudicator service handles concurrency control using an optimistic concurrency control protocol.
• The Journal service provides durable storage of transactions, replicated across multiple AZs.

Availability and Failover

• DSQL is designed for high availability, with no single points of failure. Components automatically fail over in the event of failures.
• Connections are managed by a Session Routing Layer, which can quickly provide new connections even in the event of mass connection churn.
• Applications need to handle connection errors and retries, as individual connections may fail, but the overall service remains available.

Multi-Region Resilience

• DSQL supports multi-region clusters, with a witness region to facilitate quorum-based failover between regions.
• Applications can be deployed actively in multiple regions, with a global endpoint automatically routing traffic to the closest healthy region.
• In the event of a regional failure, DSQL will automatically reconfigure to continue serving the application without data loss.

Maintenance and Upgrades

• DSQL is a fully managed service, with no maintenance required by customers.
• Automated security updates, minor version upgrades, and other maintenance tasks are handled by AWS, with no downtime or manual intervention required.

Key Takeaways

• Amazon Aurora provides built-in resilience features, including high durability storage, automated failover, and multi-region capabilities, to ensure availability and disaster recovery.
• Aurora Serverless v2 (DSQL) takes resilience a step further with a fully managed, highly available architecture that separates compute and storage, and supports active-active multi-region deployments.
• Both Aurora offerings allow customers to focus on building applications, rather than managing the underlying database infrastructure and resilience mechanisms.
• Specific technical features include:
  • 6-way data replication across 3 AZs for Aurora storage
  • Automated failover and connection management in Aurora
  • Optimistic concurrency control and quorum-based multi-region failover in DSQL
  • Fully automated maintenance and upgrades for both Aurora offerings

Business Impact

• The resilience features of Amazon Aurora and DSQL enable customers to build highly available, fault-tolerant applications that can withstand infrastructure failures and regional outages.
• This reduces the operational burden and risk associated with running mission-critical databases, allowing teams to focus on developing new features and capabilities rather than managing the underlying database platform.
• The multi-region capabilities of Aurora Global Database and DSQL's active-active architecture enable customers to serve users from the closest available region, improving application performance and responsiveness.
• The seamless maintenance and upgrade processes ensure that customers always have access to the latest database capabilities and security improvements without disrupting their applications.

Examples and Use Cases

• The presenters did not provide specific customer examples or use cases, but the resilience features of Aurora and DSQL would be beneficial for any mission-critical, globally-distributed applications that require high availability and disaster recovery, such as:
  • E-commerce platforms
  • Financial trading systems
  • SaaS applications with global user bases
  • IoT platforms that ingest and process data at scale