Cloud Cost Optimization Cheat Sheet
Practical strategies and tools for reducing cloud spend across compute, storage, and networking on AWS, Azure, and GCP.
2 PagesIntermediateFeb 10, 2026
AWS Cost Explorer CLI
Query cost and usage data from the command line.
bash
aws ce get-cost-and-usage \ --time-period Start=2026-06-01,End=2026-07-01 \ --granularity MONTHLY \ --metrics "UnblendedCost" \ --group-by Type=DIMENSION,Key=SERVICEaws ce get-rightsizing-recommendation \ --service AmazonEC2
Find Idle/Unattached Resources (AWS)
Identify unattached EBS volumes wasting spend.
bash
aws ec2 describe-volumes \ --filters Name=status,Values=available \ --query "Volumes[*].{ID:VolumeId,Size:Size,Type:VolumeType}" \ --output table
Compute Savings Levers
Key compute savings levers to know.
- Reserved Instances / Committed Use- 1-3 year commitment for 30-70% discount vs on-demand
- Savings Plans (AWS)- Flexible commitment to a $/hour spend, applies across instance families
- Spot / Preemptible Instances- Deep discounts for interruptible, fault-tolerant workloads
- Rightsizing- Matching instance size to actual CPU/memory utilization
- Auto-scaling- Scale compute down during low-traffic periods instead of running peak capacity 24/7
Storage & Monitoring Levers
Key storage & monitoring levers to know.
- Storage Lifecycle Policies- Auto-transition cold data to cheaper tiers (Glacier, Archive, Coldline)
- Delete Unattached Volumes/Snapshots- Orphaned EBS volumes and old snapshots silently accumulate cost
- Tagging Strategy- Tag resources by team/project to attribute cost and spot anomalies
- Budgets & Alerts- Set threshold-based alerts (AWS Budgets, Azure Cost Alerts) before overspend happens
- Egress Traffic- Cross-region and internet data transfer is often the hidden top cost driver
Enforce Cost Allocation Tags via Terraform
Fail the plan if required cost-tracking tags are missing, preventing untagged spend.
hcl
variable "required_tags" { type = list(string) default = ["team", "cost-center", "environment"]}check "cost_tags_present" { assert { condition = alltrue([ for t in var.required_tags : contains(keys(aws_instance.app.tags), t) ]) error_message = "Missing required cost allocation tag(s) on aws_instance.app" }}resource "aws_instance" "app" { ami = "ami-0abcdef1234567890" instance_type = "m6i.large" tags = { team = "platform" cost-center = "CC-1042" environment = "prod" }}
S3 Intelligent-Tiering + Lifecycle Policy
Automatically move objects to cheaper storage classes as access frequency drops.
json
{ "Rules": [ { "ID": "AutoTierColdData", "Status": "Enabled", "Filter": { "Prefix": "logs/" }, "Transitions": [ { "Days": 30, "StorageClass": "STANDARD_IA" }, { "Days": 90, "StorageClass": "GLACIER" }, { "Days": 365, "StorageClass": "DEEP_ARCHIVE" } ], "NoncurrentVersionExpiration": { "NoncurrentDays": 30 }, "AbortIncompleteMultipartUpload": { "DaysAfterInitiation": 7 } } ]}
Kubernetes VerticalPodAutoscaler (Recommend-Only)
Continuously right-size container requests instead of guessing at deploy time.
yaml
apiVersion: autoscaling.k8s.io/v1kind: VerticalPodAutoscalermetadata: name: api-vpaspec: targetRef: apiVersion: apps/v1 kind: Deployment name: api updatePolicy: updateMode: "Off" # recommend only; apply manually or switch to Auto once trusted resourcePolicy: containerPolicies: - containerName: '*' minAllowed: cpu: 50m memory: 64Mi maxAllowed: cpu: 2 memory: 2Gi
FinOps Metrics That Matter
Cost KPIs mature teams track beyond total spend.
- Unit Economics- Cost per request/transaction/customer — normalizes spend against actual business growth
- Committed Use Coverage- % of usage covered by Reserved Instances/Savings Plans vs on-demand — track to avoid over- or under-committing
- Effective Savings Rate- Actual discount realized after accounting for unused commitment (a poorly matched RI can net negative)
- Waste Ratio- Spend on idle/unattached/oversized resources as a % of total bill
- Showback vs Chargeback- Showback reports cost per team for awareness; chargeback actually bills it to their budget
- Anomaly Detection Lag- Time between a cost spike occurring and an alert firing — the gap is where waste accumulates
Advanced Compute Cost Levers
Beyond basic rightsizing and reserved instances.
- Graviton / ARM Migration- ARM-based instances (AWS Graviton) offer ~20-40% better price-performance for compatible workloads
- Spot Fleet Diversification- Spread Spot requests across multiple instance types/AZs to reduce interruption risk and improve fill rate
- Bin Packing / Node Consolidation- In Kubernetes, aggressive bin-packing (via Karpenter or Cluster Autoscaler) reduces the number of underutilized nodes
- Serverless for Spiky Workloads- Pay-per-invocation models beat always-on compute when traffic is bursty or intermittent
- Scheduled Scaling for Dev/Test- Auto-stop non-prod environments outside business hours; often 60-70% of a dev account's runtime is idle nights/weekends
Pro Tip
Before buying Reserved Instances or Savings Plans, run rightsizing analysis first — committing to the wrong instance size locks in waste for the entire commitment term.
Was this cheat sheet helpful?
Explore Topics
#CloudCostOptimization#CloudCostOptimizationCheatSheet#CloudComputing#Intermediate#AWS#Cost#Explorer#CLI#Networking#CheatSheet#SkillVeris
