Practise vocabulary for reserved instances, savings plans, spot instances, rightsizing, over-provisioned resources, CPU utilisation, and idle resource detection.
0 / 5 completed
1 / 5
"Rightsizing" a cloud resource means:
Rightsizing is the most common cost optimisation action: "this m5.2xlarge runs at 8% average CPU — we can downsize to m5.large with no performance impact, saving 75% on that instance." Cloud providers offer rightsizing recommendations (AWS Compute Optimizer, Azure Advisor, GCP Recommender) based on utilisation history.
2 / 5
The phrase "we're running at 8% average CPU utilisation" indicates:
Low utilisation = wasted spend. Industry benchmarks: cloud environments average 15-20% CPU utilisation on EC2. Many enterprises run at 5-10%. A $100K/month compute bill at 8% utilisation means ~$85K is wasted. Rightsizing, auto-scaling, and consolidation can recapture this. "Over-provisioned" is the vocabulary for this pattern.
3 / 5
A "spot instance" (AWS) or "preemptible VM" (GCP) is:
Spot/preemptible instances are ideal for: batch data processing, ML training jobs, CI/CD workers, and stateless web tier capacity. Pattern: use reserved/savings plan for baseline load, on-demand for burst, spot for flexible batch workloads. Teams running Spark on spot can reduce compute costs by 60-80% compared to on-demand.
4 / 5
AWS Savings Plans differ from Reserved Instances in that:
Reserved Instances are instance-type-specific (m5.large in us-east-1); switching to m5.xlarge or moving to eu-west-1 loses the discount. Savings Plans commit to $X/hour of any EC2 or Fargate usage — much more flexible. Compute Savings Plans: up to 66% off; EC2 Instance Savings Plans: up to 72% off. Choose based on workload predictability and fleet composition.
5 / 5
"Idle resource detection" in cloud cost optimisation identifies:
Idle resources are pure waste: an unattached 1TB EBS volume costs ~$80/month indefinitely. Idle RDS instances cost full price. Cloud providers accumulate these over time — unused test environments, forgotten prototypes, orphaned load balancers. Cost optimisation tooling (CloudHealth, Apptio Cloudability) scans for idle resources and flags them for termination.