Cost Optimization Review

Cost Optimization Review

Step 1: Gather context

Ask the user:

> What workload or AWS environment would you like me to review for cost optimization? Please share: > - Workload name and code packages/directories to analyze > - Traffic patterns (steady, spiky, predictable growth, seasonal) > - Monthly spend estimate (optional — helps calibrate severity) > - Budget constraints or targets (optional)

If context is already provided or you are in a codebase with IaC, proceed directly.

Step 2: Compute Cost Discovery

Analyze all compute resource configurations.

You MUST examine:

• EC2 instance types and sizes in ASG/launch templates
• ECS task definitions (CPU/memory allocations)
• Lambda memory and timeout configurations
• Fargate task sizes
• EKS node group configurations
• Batch compute environments

For each compute resource, document:

• File path and line numbers
• Instance type/size configured
• Scaling configuration (min/max/desired)
• Whether Graviton is used (or x86 where Graviton is available)
• Pricing model implications (on-demand only vs Spot-eligible vs Reserved-eligible)

You MUST flag as HIGH RISK:

• Fixed-size compute (no auto-scaling) for variable workloads
• Over-provisioned Lambda (>1024MB for simple operations)
• x86 instance types where Graviton equivalents exist (cost + sustainability)
• Large instance types for workloads that could use smaller instances with horizontal scaling
• Dev/test environments with production-sized resources
• No scheduled scaling for environments with clear off-hours

Step 3: Storage and Data Cost Discovery

Analyze storage configurations.

You MUST examine:

• S3 bucket configurations (storage class, lifecycle policies, versioning)
• EBS volume types and sizes
• RDS storage configurations (type, allocated size, auto-scaling)
• DynamoDB capacity mode and provisioning
• ElastiCache node types and cluster sizes
• EFS configurations (throughput mode, lifecycle)
• Backup retention policies (AWS Backup, RDS snapshots)
• Log retention settings (CloudWatch log groups)

You MUST flag as HIGH RISK:

• S3 buckets without lifecycle policies (accumulating indefinitely)
• CloudWatch log groups with "never expire" retention
• DynamoDB with provisioned capacity that could use on-demand (or vice versa based on pattern)
• Over-provisioned IOPS on EBS/RDS
• EBS volumes not using gp3 (gp2 is more expensive for same performance)
• Backup retention > 35 days without business justification
• S3 versioning enabled without lifecycle rules to expire old versions

Step 4: Data Transfer Cost Discovery

Analyze network and data transfer patterns.

You MUST examine:

• NAT Gateway usage (could VPC endpoints replace?)
• Cross-region data transfer patterns
• VPC endpoint configurations (or lack thereof for S3/DynamoDB)
• CloudFront distributions (or lack thereof for static content)
• Cross-AZ traffic patterns (multi-AZ services communicating across AZs)
• API Gateway configurations (REST vs HTTP API pricing)

You MUST flag as HIGH RISK:

• S3/DynamoDB access going through NAT Gateway (VPC endpoint would be free)
• No CloudFront for static content delivery
• REST API Gateway where HTTP API would suffice (70% cheaper)
• Cross-region replication without business justification
• Public S3 access through internet where CloudFront would reduce transfer costs

Step 5: Pricing Model Assessment

Analyze whether pricing models align with usage patterns.

You MUST evaluate:

• Steady-state compute → Savings Plans or Reserved Instances opportunity
• Variable/batch compute → Spot Instance opportunity
• Serverless vs provisioned alignment (Lambda/Fargate for spiky, EC2/ECS for steady)
• DynamoDB on-demand vs provisioned (on-demand for unpredictable, provisioned for steady)
• Aurora Serverless v2 vs provisioned (for variable database load)
• S3 Intelligent-Tiering for unknown access patterns

Step 6: Environment and Lifecycle Management

Analyze non-production environment configurations.

You MUST examine:

• Dev/test/staging environment sizing vs production
• Scheduled scaling or shutdown for non-production
• Resource lifecycle policies (TTL on test resources)
• Cost allocation tags on resources
• Budget and anomaly detection configurations

You MUST flag as IMPROVEMENT OPPORTUNITY:

• Non-production environments running 24/7 at production scale
• No cost allocation tags on resources
• No AWS Budget or Cost Anomaly Detection configured
• No lifecycle policies on test/temporary resources

---STOP--- Checkpoint: Discovery complete — ready to evaluate against WA Framework

> Discovered compute resources ({X} instances/functions), storage configurations ({Y} buckets/volumes/tables), data transfer patterns, pricing model alignment, and environment lifecycle settings across {packages analyzed}. > > Shall I proceed with evaluating these findings against the WA Cost Optimization pillar questions and quantifying savings estimates?

Do NOT proceed past this point until the user explicitly confirms. ---

Step 7: Evaluate against WA Framework questions

For each question, provide: Status, Evidence (file:line), Gaps, Risk.

COST 1 — How do you implement cloud financial management?

• Evidence: cost allocation tags, budget configs, anomaly detection rules

COST 2 — How do you govern usage?

• Evidence: SCPs limiting instance types, quotas, resource constraints

COST 3 — How do you monitor usage and cost?

• Evidence: Budget alarms, Cost Anomaly Detection configs, billing alarms

COST 4 — How do you decommission resources?

• Evidence: lifecycle policies, TTL configs, cleanup automation, retention rules

COST 5 — How do you evaluate cost when you select services?

• Evidence: serverless for variable loads, provisioned for steady, Spot for batch

COST 6 — How do you meet cost targets when you select resource type, size, and number?

• Evidence: instance types, scaling min/max, memory allocations, right-sizing evidence

COST 7 — How do you use pricing models to reduce cost?

• Evidence: Savings Plan configs, Reserved capacity, Spot fleet configs

COST 8 — How do you plan for data transfer charges?

• Evidence: VPC endpoints, CloudFront distributions, regional placement

Step 8: Quantify savings estimates

For each finding, estimate:

• Current cost indicator: resource configuration that drives cost (instance type, provisioned capacity, retention)
• Optimized configuration: what the resource should be
• Relative savings: percentage reduction for that resource category
• Effort: Low / Medium / High to implement
• Risk: impact on reliability or performance

Step 9: Risk Assessment

For each finding, assess using Impact × Likelihood:

Impact: Minor ( 30% savings in category or fundamental pricing model misalignment)

Likelihood (of waste continuing): Low (planned optimization exists) | Medium (no plan but low growth) | High (growing unchecked, no visibility)

| Impact | Likelihood | Risk Level | |----------|------------|------------| | Severe | High | Critical | | Severe | Medium | High | | Severe | Low | High | | Moderate | High | High | | Moderate | Medium | Medium | | Moderate | Low | Medium | | Minor | High | Medium | | Minor | Medium | Low | | Minor | Low | Low |

---STOP--- Checkpoint: Risk assessment complete — ready to produce the final report

> Assessed {N} findings across compute, storage, data transfer, pricing models, and environment management. Risk distribution: {X} Critical, {Y} High, {Z} Medium, {W} Low. Top savings opportunities identified with effort estimates. > > Shall I produce the full Cost Optimization report with prioritized remediation plan?

Do NOT proceed past this point until the user explicitly confirms. ---

Step 10: Produce the report

# Cost Optimization Review: {Workload Name}

## Executive Summary
- **Date**: {date}
- **Packages Analyzed**: {list}
- **Findings**: {X} Critical, {Y} High, {Z} Medium, {W} Low
- **Overall Cost Efficiency**: {1-5} — {one-line justification}
- **Top 3 Savings Opportunities**: {brief list}

## Cost Optimization Scorecard
| Domain | Score (1-5) | Key Strength | Key Gap |
|--------|-------------|--------------|---------|
| Compute Right-Sizing | {score} | {strength} | {gap} |
| Storage Lifecycle | {score} | {strength} | {gap} |
| Data Transfer | {score} | {strength} | {gap} |
| Pricing Models | {score} | {strength} | {gap} |
| Environment Management | {score} | {strength} | {gap} |
| Cost Visibility | {score} | {strength} | {gap} |

## Critical and High Risk Findings
{For each: ID, domain, title, description, evidence (file:line), cost impact, recommendation, effort, AWS services}

## Medium Risk Findings
{Same format, condensed}

## Low Risk Findings
{Summary table: ID | Domain | Title | Recommendation}

## Savings Summary
| Category | Finding | Current Config | Optimized Config | Relative Savings | Effort |
|----------|---------|---------------|-----------------|-----------------|--------|
| Compute | {title} | {current} | {recommended} | {%} | {effort} |
| Storage | {title} | {current} | {recommended} | {%} | {effort} |
| Transfer | {title} | {current} | {recommended} | {%} | {effort} |

## Prioritized Remediation Plan

### Quick Wins ( Would you like me to:
> - Generate IaC changes for quick-win optimizations?
> - Model Savings Plans vs Reserved Instances for your usage?
> - Design a scheduled scaling policy for non-production?
> - Create a FinOps tagging strategy with cost allocation tags?
> - Estimate cost for an architectural alternative (serverless, containers)?
> - Implement S3 lifecycle policies for identified buckets?

## Calibration Guidance

- A workload with auto-scaling, lifecycle policies, VPC endpoints, and cost allocation tags is MATURE — focus on advanced optimizations (Savings Plans, Spot, architectural alternatives)
- Every finding MUST have code evidence — don't flag "over-provisioned" without checking actual configurations
- Savings estimates should be relative (% reduction) when absolute spend is unknown
- Always assess trade-offs: cost optimization MUST NOT introduce reliability risks without explicit acknowledgment
- Flag when a cost optimization would degrade the reliability or performance posture
- "Cannot Determine" is valid when you can't assess actual utilization from IaC alone (recommend Compute Optimizer, Cost Explorer)
- Acknowledge good cost practices before listing waste

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [aws-samples](https://github.com/aws-samples)
- **Source:** [aws-samples/sample-well-architected-skills-and-steering](https://github.com/aws-samples/sample-well-architected-skills-and-steering)
- **License:** MIT-0
- **Homepage:** https://aws.amazon.com/architecture/well-architected/

Install and usage instructions live in the source repository linked above.