Cloud Cost Optimisation Interview Questions
Core Concepts
Q: What is FinOps? What are the main levers for cloud cost optimisation?
FinOps (Cloud Financial Operations) is the practice of bringing financial accountability to cloud spending — enabling engineering, finance, and business teams to make data-driven spending decisions.
FinOps phases:
1.Inform: Visibility into spending. Who spends what on what? Unit economics.
2.Optimise: Right-size, eliminate waste, use commitment discounts.
3.Operate: Continuous process. Policies, automation, governance.
Main cost levers:
| Lever | Typical Saving | Effort |
|---|
|---|---|---|
| Spot/Preemptible instances | 60-90% | Medium |
|---|
| Reserved/Committed use | 30-72% | Low (commitment required) |
| Right-sizing (CPU/memory) | 20-40% | Medium |
| Storage tiering | 50-80% for cold data | Low |
| Auto-scaling (eliminate idle) | 20-50% | Medium |
| Delete unused resources | 10-30% | Low |
| Savings Plans | 20-40% | Low (flexible commitment) |
Q: Explain commitment discounts — RI vs Savings Plans vs CUDs.
AWS:
•Reserved Instances (RI): Commit to specific instance type/region for 1-3 years. 40-72% off.
•Savings Plans (Compute): Commit to $/hour of compute spend (flexible across types). 17-66% off.
•Savings Plans (EC2): Commit to specific instance family. Up to 72% off.
Azure:
•Reserved VM Instances: 1-3 year commit, specific VM size/region. 40-72% off.
•Azure Hybrid Benefit: Bring your own Windows Server/SQL Server license.
GCP:
•Committed Use Discounts (CUDs): 1-3 year commit for CPU/RAM or specific resources.
•Sustained Use Discounts: Automatic discounts for VMs running >25% of month (up to 30%).
Right-sizing before committing: Never buy RIs before right-sizing — you lock in waste.
Q: How do you identify and eliminate cloud waste?
Step 1: Discover waste
bash
# AWS: Trusted Advisor, Cost Explorer, Compute Optimizer
aws ce get-cost-and-usage --time-period Start=2024-01-01,End=2024-01-31 --granularity MONTHLY --group-by Type=DIMENSION,Key=SERVICE
# AWS Compute Optimizer: right-sizing recommendations
aws compute-optimizer get-ec2-instance-recommendations
# Azure Advisor: cost recommendations
az advisor recommendation list --category Cost
Common waste patterns:
•Idle instances: Low CPU (<5%) for extended periods → downsize or terminate
•Unattached EBS/disks: Volumes not attached to any instance
•Unassociated Elastic IPs: Charged when not attached ($3.60/month each)
•Old snapshots: Auto-delete snapshots older than 30 days (lifecycle policy)
•Dev/test running 24/7: Schedule to run 8 hours/day → save 67%
•Over-provisioned RDS: Memory utilisation <20% → downsize
•Unused NAT Gateways: Charged per GB + hourly (~$32/month + data)
•Data transfer costs: Cross-AZ, cross-region, internet egress
Q: Cloud cost allocation and showback/chargeback.
Tagging strategy (foundation of cost allocation):
Mandatory tags:
- Project: my-project
- Environment: production/staging/dev
- Team: backend/frontend/data
- CostCenter: 1234
- Owner: alice@company.com
Enforce with:
AWS: Tag Policies in AWS Organizations
Azure: Azure Policy (deny resources without required tags)
GCP: Resource Manager labels + org policies
Showback: Show teams what they spend (no charge).
Chargeback: Actually charge teams/departments for their cloud usage.
Unit economics: Cost per transaction, cost per user, cost per API call.
Good metric to track: infrastructure cost as % of revenue (target <5% for SaaS).
Q: Auto-scaling for cost efficiency.
yaml
# K8s HPA: scale pods based on metrics
apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
spec:
minReplicas: 1 # Scale to zero when possible
maxReplicas: 20
metrics:
- type: Resource
resource:
name: cpu
target:
type: Utilization
averageUtilization: 70
# KEDA: scale to zero on queue depth
spec:
triggers:
- type: aws-sqs-queue
metadata:
queueURL: https://sqs.us-east-1.amazonaws.com/123/my-queue
queueLength: "5" # Scale up when >5 messages
minReplicaCount: 0 # Scale to zero when queue empty
Scheduled scaling: Scale down dev environments at night/weekends.
Revision Notes
FINOPS: Inform (visibility) → Optimise (right-size, commitments) → Operate (governance)
COST LEVERS (highest impact first):
1. Spot/Preemptible: 60-90% off (stateless/batch)
2. Reserved/CUD: 30-72% off (1-3yr commit) — right-size BEFORE committing
3. Right-sizing: 20-40% (Compute Optimizer, Advisor)
4. Storage tiering: 50-80% for cold data
5. Delete waste: idle instances, unattached volumes, unused IPs, old snapshots
TAGGING: Project + Environment + Team + CostCenter + Owner
Enforce via Tag Policies / Azure Policy / GCP org labels
Showback (visibility) → Chargeback (billing by team)
COMMITMENT TYPES:
AWS: Reserved Instances (specific) or Savings Plans (flexible)
Azure: Reserved VM Instances + Hybrid Benefit (BYOL)
GCP: CUDs + Sustained Use Discounts (automatic)
WASTE DETECTION:
CPU <5% → downsize | Unattached EBS → delete | Dev 24/7 → schedule 8hr/day
Old snapshots → lifecycle policy | Unused NAT GW → consolidate
