Reducing RDS Costs: Aurora and Read Replicas

RDS Cost Optimization Strategies

Amazon RDS is often the single largest line item after EC2. Database instances run 24/7, storage costs accumulate, and backup retention adds up. Here are proven strategies to cut RDS costs without sacrificing performance or availability.

Strategy 1: Right-Size RDS Instances

Most RDS instances are over-provisioned. Check these CloudWatch metrics to identify right-sizing opportunities:

• CPUUtilization — If p95 is below 30%, downsize the instance class
• FreeableMemory — If consistently above 50% of total memory, you have headroom
• DatabaseConnections — Peak connections determine the minimum instance size

Strategy 2: Aurora Serverless v2

For variable workloads, Aurora Serverless v2 scales capacity in 0.5 ACU increments (1 ACU ≈ 2 GB RAM). You pay only for the capacity you use, down to 0.5 ACU during idle periods.

Best for: Dev/test databases, applications with unpredictable traffic, overnight batch processing windows.

Cost example: A database that peaks at 8 ACUs for 4 hours/day and idles at 1 ACU for 20 hours costs ~60% less than a provisioned db.r6g.xlarge running 24/7.

Strategy 3: Read Replicas for Read-Heavy Workloads

Instead of scaling up your primary instance, add read replicas to handle SELECT queries:

• Route read-heavy analytics queries to replicas
• Use Aurora's reader endpoint for automatic load balancing across replicas
• Use smaller instance types for replicas if they handle lighter workloads

Strategy 4: Reserved Instances for RDS

RDS Reserved Instances offer 30-60% savings over On-Demand pricing. Since databases are typically long-running, they're excellent candidates for 1-year or 3-year commitments.

Strategy 5: Storage and Backup Optimization

1. Use gp3 storage instead of io1 unless you need guaranteed high IOPS
2. Reduce backup retention to the minimum required (7 days for dev, 35 for production)
3. Delete manual snapshots that are no longer needed
4. Enable storage auto-scaling to avoid over-provisioning upfront

Real-world result: Moving a production PostgreSQL workload from db.r5.2xlarge to Aurora Serverless v2 with read replicas reduced monthly RDS costs from $2,100 to $780.