Cost-Optimized Databases

Bundled into the existing AWS Certified Solutions Architect – Associate premium course — no separate purchase.

Included in this chapter:

Service	Commitment mechanism	Variable-load mechanism	Caching layer benefit
RDS (MySQL, PostgreSQL, etc.)	RDS Reserved Instances (1/3 yr)	Stop instance up to 7 days	ElastiCache for read offload
Aurora	Aurora Reserved Instances	Aurora Serverless v2 (ACU-based autoscaling)	ElastiCache for hot keys
DynamoDB	Reserved Capacity (provisioned mode)	On-demand mode	DAX for read-through cache
ElastiCache (Redis, Memcached, Valkey)	Reserved Nodes (1/3 yr)	Serverless mode (ECPU-based)	N/A (it's the cache)
Redshift	Reserved Nodes (1/3 yr)	Serverless (RPU-based)	Result-set cache built-in

Cheat sheet

Commit capacity for predictable load, pay per use for spiky load
Put a cache in front of a saturating read-heavy DB instead of upsizing it
DynamoDB: provisioned for steady high use, on-demand for spiky
Aurora Serverless v2 only wins below ~50% utilization
ElastiCache Reserved Nodes use the same RI payment options as EC2
Stop an idle RDS instance to drop compute cost (max 7 days)
Lazy-load by default, use write-through when a miss is catastrophic
Switch Aurora to I/O-Optimized once I/O exceeds ~25% of cost
Use ElastiCache Serverless for unpredictable cache traffic (Valkey 33% cheaper)
Use ElastiCache R6gd data tiering when ≤20% of the dataset is hot (60%+ savings)
Customer-managed KMS keys cost ~$1/key/month, so consolidate where you can
Aurora Serverless v2 scales 0–256 ACUs in 0.5-ACU steps

Unlock with Premium — includes all practice exams and the complete study guide.

Cheat sheet