Exam relevance: Domain 3 (25%). Cost Explorer is the primary tool for analyzing and forecasting AWS costs. Know filtering, grouping, rightsizing recommendations, and Savings Plans analysis.

1. Overview

AWS Cost Explorer provides interactive visualizations and analysis of AWS costs and usage. It helps identify cost drivers, forecast future spend, and find optimization opportunities.

2. Key Features

Cost and Usage Analysis

  • Visualize costs by: service, account, region, tag, instance type, purchase option
  • Filter by: time range (up to 13 months historical), service, linked account, tag
  • Group by multiple dimensions simultaneously
  • Daily, monthly, or hourly granularity (hourly requires additional cost)

Forecasting

  • ML-based cost forecast for up to 12 months
  • Based on historical usage patterns
  • Confidence intervals included
  • Use for: budget planning, capacity planning

Rightsizing Recommendations

  • Identifies over-provisioned EC2 instances
  • Recommends instance type changes based on CloudWatch utilization metrics
  • Shows estimated monthly savings
  • Requires Cost Explorer rightsizing recommendations enabled (free)
  • Considers: CPU, memory (requires CloudWatch Agent), network utilization

Savings Plans Recommendations

  • Recommends Compute, EC2 Instance, or SageMaker Savings Plans
  • Based on historical On-Demand usage
  • Shows: recommended commitment, estimated savings, coverage
  • Lookback period: 7, 30, or 60 days

3. Cost Allocation Tags

  • Tag resources to track costs by project, team, environment, cost center
  • AWS-generated tags: aws:createdBy (who created the resource)
  • User-defined tags: custom key-value pairs
  • Activate tags in Cost Explorer for cost allocation
  • Tags appear in Cost Explorer after 24 hours

4. Budgets Integration

  • Create budgets based on Cost Explorer data
  • Alert when actual or forecasted costs exceed thresholds
  • Budget types: cost, usage, Savings Plans coverage, reservation coverage
  • Actions: notify via SNS, apply IAM policy, apply SCP (to restrict spending)

5. Exam Scenarios

Scenario 1: A company wants to identify which AWS accounts and services are driving the most cost increase over the last 3 months. What is the solution?

Use Cost Explorer. Filter by the last 3 months. Group by Linked Account and Service. Sort by cost to identify the top cost drivers. Drill down into specific accounts or services for further analysis.

Scenario 2: A company wants to reduce EC2 costs. They suspect many instances are over-provisioned. What is the solution?

Enable Cost Explorer Rightsizing Recommendations. Install the CloudWatch Agent on instances to collect memory metrics. Cost Explorer analyzes CPU, memory, and network utilization and recommends smaller instance types with estimated savings.

Scenario 3: A company has steady EC2 usage and wants to commit to Savings Plans but doesn’t know the right commitment amount. What is the solution?

Use Cost Explorer Savings Plans Recommendations. Set the lookback period to 30 days. Cost Explorer recommends the optimal commitment amount based on historical On-Demand usage, showing estimated savings and payback period.

6. Exam Tips

  • Cost Explorer shows 13 months of historical data
  • Rightsizing recommendations require CloudWatch metrics — memory needs CloudWatch Agent
  • Cost allocation tags must be activated in Cost Explorer before they appear in reports
  • Cost Explorer is free to use; hourly granularity has an additional cost
  • Savings Plans recommendations are based on historical usage — use 30-day lookback for stable workloads
  • For real-time cost alerts, use AWS Budgets (Cost Explorer is for analysis, not alerting)

7. 🎯 Architecture Decision Notes

Cost optimization starts at architecture time

Treat cost as a design constraint, not a spreadsheet exercise at the end:

  • Match commitment instruments to stable baselines: Savings Plans and Reserved Instances discount predictable usage; they do not make spiky workloads cheap — right-sizing and serverless do.
  • Savings Plans vs RIs: Plans flex across instance families, regions, and services (Compute plans cover EC2, Lambda, Fargate); RIs lock to configuration. Flexibility has a price — model both.
  • Tag-driven allocation only works with a governed tag vocabulary enforced by policy; untagged resources make every cost report fiction.
  • The cheapest architecture that misses RTO/RPO is not cheap — it is just unpriced risk.

8. Official References