Cost Optimization and Performance
What the exam tests
Cost Optimization and Performance tests rightsizing, pricing models, storage lifecycle, elasticity, and performance efficiency. Expect tagging, budgets, savings plans or reserved instances, Spot, CDN caching, and database tuning.
Core rules
- Make cost visible: enforce a tagging schema for owner, environment, and cost center, then review untagged spend.
- Eliminate waste first: idle instances, unattached volumes, old snapshots, and over-provisioned databases.
- Match pricing to workload: on-demand for spiky, savings plans or reserved capacity for steady state, Spot for fault-tolerant batch.
- Optimize performance with caching, CDNs, read replicas, and asynchronous processing before scaling up.
Worked example
CloudWatch shows an EC2 fleet averaging 12% CPU over 30 days. Rightsize from m5.2xlarge to m5.large, then purchase a one-year compute savings plan for the steady baseline and move batch jobs to Spot. Reasoning: rightsizing cuts unit cost immediately; the savings plan discounts predictable usage; Spot absorbs interruption-tolerant work at a steep discount. Recheck latency after two weeks to confirm no regression.
Traps
Buying reserved capacity before rightsizing. Ignoring data-transfer and NAT gateway costs. Leaving dev and test environments running nights and weekends. No budgets or anomaly alerts. Scaling vertically when a cache or queue would fix the bottleneck.