Amazon Aurora

Definition

Amazon Aurora is a cloud-native relational database engine built by AWS, fully compatible with MySQL and PostgreSQL. Aurora redesigns the database storage layer to run distributed across 6 storage nodes in 3 Availability Zones — delivering up to 5x the throughput of standard MySQL and 3x PostgreSQL performance, with higher availability and lower storage costs than equivalent self-managed or standard RDS deployments.

How Aurora is Different from Standard RDS

Distributed Storage Architecture:

• Data is replicated across 6 storage nodes in 3 AZs — writes go to all 6; reads require 3/6 agreement (quorum)
• Storage auto-scales from 10 GB to 128 TB in 10 GB increments — no pre-provisioning
• Storage is separate from compute: failover is faster because standby instances attach to the same shared storage

Performance:

• MySQL-compatible: up to 5x standard MySQL throughput
• PostgreSQL-compatible: up to 3x standard PostgreSQL throughput
• Parallel query: offloads analytical queries to the storage layer without impacting OLTP workloads
• Up to 15 low-latency read replicas (vs 5 for standard RDS)

Availability:

• 99.99% uptime SLA (vs 99.95% for Multi-AZ RDS)
• Automatic failover in under 30 seconds (vs 60–120 seconds for standard RDS Multi-AZ)
• Write forwarding: read replicas forward writes to the primary automatically

Aurora Serverless v2

Aurora Serverless v2 scales compute capacity instantly based on workload demand:

• Scales in fine-grained increments (0.5 ACU steps) — no hard tier jumps
• Scales from near-zero to hundreds of ACUs within milliseconds
• Can be mixed with provisioned instances in the same Aurora cluster
• Per-second billing when active; minimal cost when idle
• Best for: variable workloads, dev/test, SaaS with diverse customer activity patterns

ACU (Aurora Capacity Unit): ~2 GB RAM + proportional CPU and network

Aurora Global Database

Aurora Global Database enables cross-region replication with near-zero lag:

• Primary region handles writes; up to 5 secondary regions serve local reads
• Typically under 1 second replication lag globally
• Secondary region can be promoted to primary in under 1 minute (planned failover) or ~1 minute (unplanned)
• Used for: globally distributed applications, cross-region disaster recovery, read latency reduction for international users

Aurora vs Standard RDS

Aurora Machine Learning Integration

Aurora integrates natively with Amazon Bedrock and SageMaker:

• Call ML models from SQL using aws_bedrock_invoke_model() and aws_sagemaker_invoke_endpoint() functions
• Run sentiment analysis, fraud detection, or product recommendations directly from your SQL queries
• No data movement — inference happens within your Aurora instance

Common Mistakes

Mistake 1: Using Aurora for small, low-traffic workloads. Aurora has a minimum cost floor higher than db.t3.micro RDS. For dev/test or very low-traffic applications, standard RDS is more cost-effective.

Mistake 2: Not enabling Aurora Serverless v2 for variable workloads. Fixed-size Aurora instances waste capacity during low-traffic periods. Serverless v2 eliminates over-provisioning while maintaining instant scale-out.

Mistake 3: Treating Aurora as a drop-in MySQL/PostgreSQL replacement without testing. Aurora has minor behavioral differences in replication, locking, and some edge-case query behaviors. Always run a compatibility test suite before migrating.

Related AWS Services

• Amazon RDS: Standard managed MySQL, PostgreSQL, Oracle, SQL Server (see Amazon RDS)
• Aurora Machine Learning: Native SQL integration with Bedrock and SageMaker
• AWS DMS: Migrate existing databases to Aurora with minimal downtime
• AWS Secrets Manager: Auto-rotate Aurora credentials

Related FactualMinds Content

• AWS RDS Consulting
• AWS Database Migration Strategy — Oracle and SQL Server to Aurora PostgreSQL with DMS Schema Conversion
• AWS Data Analytics
• Cloud Compliance Services