Cloud Cost Optimization Language | Cloud FinOps Exercises

1 / 5

An engineer suggests: "We should rightsize the EC2 instances — several are running at under 10% CPU."

What does rightsizing mean?

Rightsizing is data-driven: analyse actual CPU/memory utilisation (target: 60–80%), then select the appropriate smaller instance.

Tool	What it does
AWS Compute Optimizer	Analyses 14 days of CloudWatch metrics using ML; provides risk-rated resize recommendations
Azure Advisor	Analyses VM utilisation; recommends resize or shutdown with estimated savings

Key vocabulary: "rightsizing", "instance utilisation", "overprovisioning", "Compute Optimizer", "right-size recommendation".

2 / 5

The team is evaluating Reserved Instances (RIs) vs. Savings Plans for their production infrastructure.

What is the key difference between these two commitment-based discount models?

Both offer significant discounts in exchange for commitment — the trade-off is flexibility vs. discount depth.

Type	Flexibility	Discount
Reserved Instance	Low (specific type/region)	Up to 72%
Savings Plan	High ($/hr commitment)	Up to 66%

Key vocabulary: "Reserved Instance (RI)", "Savings Plan", "commitment-based discount", "convertible RI".

3 / 5

The team is considering spot instances for their nightly batch processing workload.

What is the primary risk of using spot instances?

Spot instance discount: up to 90% vs. on-demand — in exchange for accepting the possibility of sudden termination.

Workload type	Spot-suitable?	Reason
Batch ML training	✓ Excellent	Checkpointing to S3; job resumes from last checkpoint
CI/CD build agents	✓ Good	Failed builds restart; no user-visible impact
Customer-facing API servers	✗ Unsuitable	2-min termination = dropped requests, session loss

Key vocabulary: "spot instance", "interruption notice", "fault-tolerant workload", "spot interruption handler", "mixed fleet (on-demand + spot)".

4 / 5

A FinOps review identifies data egress costs as the top spend item — larger than compute.

What are data egress costs?

Data transfer costs are one of the most common surprises in cloud billing, as they are often overlooked during architecture design.

Transfer type	AWS cost
Egress to internet	~$0.09/GB (most expensive; most common surprise)
Inter-region transfer	~$0.02/GB each direction
Same-region, same-AZ	Free

Architecture implications: co-locate services in the same AZ, use CDNs for static assets, avoid unnecessary cross-region calls. Key vocabulary: "data egress", "data transfer costs", "inter-region transfer", "CDN offload", "network cost architecture".

5 / 5

The platform team is enforcing tag coverage across all cloud resources.

Why is tag coverage important for FinOps?

Untagged spend = dark matter in FinOps. If you cannot attribute a cost, you cannot manage, optimise, or hold anyone accountable for it.

Tag coverage	Effect
95%+	Reliable showback/chargeback; small unallocated bucket
70–80%	Significant unallocated spend; cost anomalies can hide in it
<50%	Cost allocation reports unreliable; FinOps programme fails

Key vocabulary: "tag coverage", "unallocated spend", "tagging policy enforcement", "tag inheritance", "dark matter spend".