Amazon Aurora

Aurora is MySQL and PostgreSQL compatible with a distributed, self-healing storage layer that replicates data across 6 storage nodes in 3 AZs. It delivers up to 5x throughput of standard MySQL and 3x throughput of standard PostgreSQL at 1/10th the cost of commercial databases.

How Aurora Works

Architecture

┌─────────────────────────────────────────────────────────────┐
│  Aurora Cluster                                              │
│                                                              │
│  ┌─────────────┐   ┌─────────────┐   ┌─────────────┐       │
│  │   Writer    │◄──│  Reader 1   │   │  Reader 2   │       │
│  │  (primary)  │   │ (async rep) │   │ (async rep) │       │
│  └──────┬──────┘   └──────┬──────┘   └──────┬──────┘       │
│         │                 │                 │               │
│         ▼                 ▼                 ▼               │
│  ┌──────────────────────────────────────────────────────┐  │
│  │           Aurora Distributed Storage Layer             │  │
│  │                                                        │  │
│  │   AZ-1          AZ-2          AZ-3                     │  │
│  │  ┌──────┐    ┌──────┐    ┌──────┐                    │  │
│  │  │ Node 1│    │ Node 2│    │ Node 3│                   │  │
│  │  └──────┘    └──────┘    └──────┘                    │  │
│  │  ┌──────┐    ┌──────┐    ┌──────┐                    │  │
│  │  │ Node 4│    │ Node 5│    │ Node 6│                   │  │
│  │  └──────┘    └──────┘    └──────┘                    │  │
│  └──────────────────────────────────────────────────────┘  │
└─────────────────────────────────────────────────────────────┘

Data is written to 6 of 6 nodes (quorum: 4 of 6). Data is read from 4 of 6 nodes.

Aurora MySQL vs Aurora PostgreSQL

	Aurora MySQL	Aurora PostgreSQL
MySQL compatible	5.7, 8.0	N/A
PostgreSQL compatible	N/A	13, 14, 15, 16
Parallel query	Yes	No
Backtrack	Yes	No
Serverless v2	Yes	Yes
Global Database	Yes	Yes
ML integrations	Yes	Yes

Endpoints

Aurora has multiple endpoints for different use cases:

Endpoint	Use	Points To
Cluster endpoint	Primary writes	Writer DB instance
Reader endpoint	Read-only queries	All reader instances (load-balanced)
Custom endpoint	Specific instance(s)	Named instances
Instance endpoint	Specific instance	Single instance

# Get all endpoints
aws rds describe-db-clusters \
  --db-cluster-identifier my-aurora-cluster \
  --query 'DBClusters[0].{ClusterEndpoint:Endpoint,ReaderEndpoint:ReaderEndpoint}'

Creating an Aurora Cluster

# Create Aurora cluster (serverless v2)
aws rds create-db-cluster \
  --db-cluster-identifier my-aurora-cluster \
  --engine aurora-postgresql \
  --engine-version 15.3 \
  --serverless-v2-scaling-configuration '{
    "MinCapacity": 1,
    "MaxCapacity": 16
  }' \
  --master-username postgres \
  --master-user-password SecretPassword \
  --vpc-security-group-ids sg-xxxxx \
  --db-subnet-group my-subnet-group \
  --backup-retention-period 1
 
# Add reader
aws rds create-db-instance \
  --db-instance-identifier my-aurora-reader \
  --db-cluster-identifier my-aurora-cluster \
  --db-instance-class db.serverless \
  --engine aurora-postgresql

Aurora Serverless v2

Aurora Serverless v2 scales ACU (Aurora Capacity Units) automatically:

# Create serverless v2 cluster
aws rds create-db-cluster \
  --db-cluster-identifier my-aurora-serverless \
  --engine aurora-postgresql \
  --serverless-v2-scaling-configuration '{
    "MinCapacity": 0.5,
    "MaxCapacity": 32
  }' \
  --master-username postgres \
  --master-user-password SecretPassword
 
# Scale manually
aws rds modify-db-cluster \
  --db-cluster-identifier my-aurora-serverless \
  --serverless-v2-scaling-configuration '{
    "MinCapacity": 1,
    "MaxCapacity": 64
  }'

ACU Pricing

	Aurora Serverless v2
Per ACU-hour	$0.12/hr
Storage	$0.10/GB/month
I/O	$0.20 per million requests

Global Database (Cross-Region)

Aurora Global Database replicates to secondary regions with < 1 second lag:

# Create global database
aws rds create-global-cluster \
  --global-cluster-identifier my-global-cluster \
  --engine aurora-postgresql \
  --engine-version 15.3
 
# Add secondary region cluster
aws rds create-db-cluster \
  --db-cluster-identifier my-aurora-secondary \
  --global-cluster-identifier my-global-cluster \
  --engine aurora-postgresql \
  --db-subnet-group-name my-subnet-group

Use case: Primary in us-east-1, secondary in eu-west-1. Failover: promote secondary to primary in < 1 second.

Backtrack (Aurora MySQL only)

Rewind the database to a specific point in time (up to 72 hours):

# Backtrack to 1 hour ago
aws rds backtrack-db-cluster \
  --db-cluster-identifier my-aurora-mysql \
  --backtrack-to 2024-01-15T09:00:00Z

This is instant (unlike restore from backup). Useful for “oops, I deleted the wrong table.”

High Availability

Multi-Master

All instances can write (MySQL only):

aws rds create-db-cluster \
  --db-cluster-identifier my-aurora-multimaster \
  --engine aurora-mysql \
  --engine-version 8.0 \
  --serverless-v2-scaling-configuration '...'

Fault Tolerance

• 6 storage nodes across 3 AZs
• Quorum: 4 of 6 writes must succeed
• Automatic failover to reader: < 30 seconds
• No data loss on single-AZ failure

Performance

Metric	Aurora MySQL	Aurora PostgreSQL	Standard MySQL
Throughput	5x MySQL	3x PostgreSQL	Baseline
Max connections	160,000	65,535	151 (default)
Max storage	128 TB	128 TB	64 TB

Monitoring

# Aurora metrics
aws cloudwatch list-metrics \
  --namespace AWS/RDS \
  --metric-name ServerlessDatabaseCapacity

Key Aurora-specific metrics:

• ServerlessDatabaseCapacity — current ACU usage
• AuroraVolumeBytesChanged — data written to storage
• RollbackSegmentInflation — undo header inflation
• FreeLocalStorage — temp storage available

Limits

Resource	Limit
Max storage	128 TB
Max instances per cluster	1 writer + 15 readers
Max connections	160,000 (Aurora MySQL), 65,535 (Aurora PG)
Backtrack window	72 hours (MySQL only)
Global database regions	5 secondary regions

References

• Homepage: https://aws.amazon.com/rds/aurora/
• Documentation: https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/
• Pricing: https://aws.amazon.com/rds/aurora/pricing/

Pricing Examples

Scenario 1: An Aurora PostgreSQL serverless v2 cluster (1-16 ACU) running an API with variable load. Average: 2 ACU, 16hr/day. 2 ACU × $0.12\times 16hr\times 30days=$115.20/month. Storage 500GB × $0.10=$50/month. Total: ~$165/month.ComparetoRDSdb.r6g.largeMulti-AZ(2\times$0.252 = $0.504/hr\times 24\times 30=$362/month). Aurora is 54% cheaper for variable workloads.

Scenario 2: A high-traffic e-commerce site needing consistent 8 ACU. 8 ACU × $0.12/hr\times 24\times 30=$691.20/month. Reserved ACU (1 year): 8 ACU × $0.075/hr=$0.60/hr × 24 × 30 = $432/month.Storage2TB\times$0.10 = $200/month.Total:$632/month. Compare to RDS db.r6g.2xlarge × 2 Multi-AZ: 2 × $0.504/hr\times 24\times 30=$725/month.

Nuggets & Gotchas

• Aurora Serverless v2 has a cold start delay (~30 seconds) when scaling from 0 ACU: If you scale to 0 during quiet periods, the first request after idle has a 30-second delay. Use MinCapacity: 0.5 to keep some capacity warm.
• Aurora MySQL backtrack requires the InnoDB tablespace to be at least 128MB: If your tables are small, you may get a “tablespace too small” error. You can increase it with innodb_undo_tablespaces.
• Aurora Global Database allows one secondary region to have a writable instance (promoted primary): In a global database, only one region is writable at a time. If you need to write in multiple regions simultaneously, use Aurora Multi-Master (MySQL only) instead.
• Aurora’s reader endpoint load-balances at the connection level, not the query level: Each new connection goes to a different reader. For true query-level load balancing, use a connection pooler (like PgBouncer for PostgreSQL or ProxySQL for MySQL).
• Aurora PostgreSQL’s shared_buffers parameter should be set to 75% of Aurora’s buffer cache, not the instance’s memory: Aurora’s storage layer is separate from PostgreSQL’s shared_buffers. The default shared_buffers (128MB) is fine for most Aurora workloads — don’t blindly set it to 75% of instance memory.