Cost Explorer granularity options là gì?

Hourly (data retained 14 ngày), Daily (38 tháng), Monthly (38 tháng). Hourly granularity có phí thêm nếu dùng API nhưng miễn phí trong console. Hourly hữu ích để identify spikes trong ngày. Monthly đủ cho high-level budget review. Daily cho trend analysis theo tuần. Cũng có thể filter theo Service, Linked Account, Tag, Region, Usage Type, Instance Type.

Cost Anomaly Detection có miễn phí không?

Miễn phí — Cost Anomaly Detection (2020) không có phí riêng. Sử dụng ML để detect cost anomalies so với historical patterns, gửi alert qua email hoặc SNS khi phát hiện. Có thể setup monitors theo individual service, AWS account, cost category, hoặc cost allocation tag. Phát hiện được ransomware mining, data exfiltration, code bugs gây cost spike — trước khi nhận bill cuối tháng.

Cost Allocation Tags hiện trong Cost Explorer sau bao lâu?

24 giờ sau khi activate. Điều quan trọng: chỉ cost data sau thời điểm activate mới được tag — không retroactive. Và tags phải được activate trong Billing Console (Billing → Cost Allocation Tags) — chỉ tag resource không đủ. Workflow: tag resources → activate tag → đợi 24h → filter theo tag trong Cost Explorer.

CUR khác Cost Explorer ở điểm gì?

Cost and Usage Report (CUR): raw, granular CSV/Parquet data export sang S3 — mọi line item (hourly), resource IDs, pricing details, reservation usage. Dùng với Athena, QuickSight, third-party tools cho custom analysis. Cost Explorer: interactive visualization, predefined dashboards, forecasting, recommendations — không granular như CUR nhưng dễ dùng hơn. CUR cho advanced analytics; Cost Explorer cho quick insights.

Compute Optimizer khác Cost Explorer recommendation ở điểm gì?

Compute Optimizer: dùng ML analyze CloudWatch utilization metrics (CPU, network, memory từ agent) để recommend optimal EC2 instance type, Lambda memory, EBS size — rất cụ thể với performance trade-offs. Cost Explorer right-sizing: recommend based on CloudWatch metrics đơn giản hơn, focus trên cost savings của EC2. Compute Optimizer có ML model tốt hơn, analyze Lambda và EBS — Cost Explorer chỉ EC2.

Tuần 10 - Ngày 1: Cost Explorer và AWS Budgets

Mục tiêu học tập

Phân tích chi phí với Cost Explorer
Setup AWS Budgets với alerts
Sử dụng Cost Anomaly Detection
Apply Cost Allocation Tags

1. AWS Cost Management Suite

Tổng quan các tools

AWS Billing Console: hóa đơn, thanh toán
Cost Explorer: visualize + phân tích chi phí
AWS Budgets: set alerts về cost/usage
Cost Anomaly Detection: ML detect anomalies
AWS Cost Categories: group costs by custom logic
AWS Pricing Calculator: estimate before deploy
Cost & Usage Report (CUR): detailed CSV/Parquet

2. Cost Explorer (Recap + Advanced)

Visualize options

Granularity: Hourly (14 days history), Daily (38 months), Monthly (38 months)
Group by: Service, Region, Account, Tag, Usage Type, Instance Type, Linked Account, Charge Type, Purchase Option
Filter: similar dimensions
Forecast: predict next 1-12 months based on past

Reports

Daily, Monthly, Year-over-Year
Save filter as Custom Report
Schedule email reports (Cost Reports)

Reserved Instance / Savings Plans Recommendations

Analyze past usage
Suggest RI/SP purchase for savings
Show estimated savings

Right-sizing recommendations

Identify underutilized EC2
Suggest smaller instance type
Combined với Compute Optimizer (more detailed)

3. AWS Budgets

4 budget types

Cost budget: dollars
Usage budget: hours, GB, requests
RI/SP budget: utilization, coverage
Action-based budget: trigger actions on threshold

Actions (Budget Actions)

Send SNS notification
Apply IAM policy (restrict permissions)
Stop EC2/RDS instances
Target SCP in Organizations

Alerting

Email (up to 10 recipients)
SNS topic
AWS Chatbot (Slack, Teams)

Cost

2 budgets free
$0.02/budget/day after

Monitor types:
- AWS services (per service)
- Linked Account (per account in Organization)
- Cost Categories
- Cost Allocation Tags
Threshold: dollar amount OR percentage increase
Subscription: email/SNS

Detection

Algorithm learns normal pattern (14+ days)
Alerts on anomalies (statistical significance)
Suggests root cause (which service/resource increased)

Cost

Free

Use case

Detect hacker mining crypto on EC2
Catch app bug causing Lambda invocation spike
Identify cost surprises before month-end

5. Cost Allocation Tags

2 types

AWS-generated: aws:createdBy, aws:cloudformation:stack-name
User-defined: Project, Environment, CostCenter, Owner

Activation

Tag resources (manual hoặc via Tag Policies)
Billing Console → Cost Allocation Tags → Activate
Wait ~24 hours for tag to appear in Cost Explorer
Filter/group by tag in Cost Explorer

Best practice

Mandatory tags via SCP: enforce tagging
Tag policies via Organizations: standardize values
Tag missing → automation: Lambda finds untagged → notify owner

Common tag schema

Project       (string): "ProjectX"
Environment   (enum): "prod", "staging", "dev", "test"
Owner         (email): "alice@company.com"
CostCenter    (string): "CC-1234"
Application   (string): "billing-service"

Define category with rules:
- Match by account, service, tag, charge type
- Multiple categories per cost

Example

Use case

Multi-team chargeback
Department billing
Project P&L

CSV, Parquet, ORC
Includes: every line item with usage, cost, tax, credits, etc.

Analysis

Athena: SQL queries
QuickSight: visualizations
3rd party tools: Cloudability, CloudHealth

vs Cost Explorer

Cost Explorer: visual UI, summarized
CUR: raw data, granular, customizable

Use case

Custom dashboards beyond Cost Explorer
FinOps team analysis
Chargeback to departments

8. AWS Pricing Calculator

Workflow

Add services to estimate (EC2, S3, etc.)
Configure: instance type, hours, storage, transfer
Calculator estimates monthly + 12-month cost
Export PDF/CSV
Share via URL

Use case

Estimate before deploy
Present to stakeholders / management
Compare options (EC2 vs Lambda vs Fargate)
Migration planning

URL

https://calculator.aws/

9. Cost Optimization Workflow

Monthly cycle

Review Cost Explorer (last month)
Identify top cost drivers, anomalies
Right-size with Compute Optimizer
Reserved Instance / Savings Plans recommendations
Tag compliance check (untagged resources)
Budget review vs actual
Action items: optimization, cleanup

Quarterly

Audit unused resources (idle EC2, unattached EBS)
Review storage classes (S3 lifecycle)
Check Reserved Instance utilization
Renew or modify RI as needed

10. Common Cost Optimization Strategies

1. Right-sizing

Compute Optimizer recommendations
Schedule shutdowns for dev/test (off-hours)
AWS Instance Scheduler service

2. Reserved Instances / Savings Plans

3-year AURI: up to 72% off
Compute Savings Plans: flexible

3. Spot Instances

Fault-tolerant workloads
Up to 90% off

4. Storage tiering

S3 lifecycle to IA/Glacier
EBS gp2 → gp3 (~20% saving)
EFS Lifecycle Management

5. Reduce data transfer

VPC Endpoints (avoid NAT/IGW)
CloudFront caching
Same-region traffic

6. Delete unused resources

Idle EC2, RDS, ELB
Unattached EBS volumes
Unused EIPs
Old snapshots

7. Lambda vs EC2 for low traffic

Lambda free tier: 1M requests/month
Often cheaper for sporadic workload

8. Use ARM (Graviton)

t4g, m6g, c7g instances ~20% cheaper

Câu hỏi ôn tập

Cost Explorer granularity options là gì?

Xem đáp án

Hourly (data retained 14 ngày), Daily (38 tháng), Monthly (38 tháng). Hourly granularity có phí thêm nếu dùng API nhưng miễn phí trong console. Hourly hữu ích để identify spikes trong ngày. Monthly đủ cho high-level budget review. Daily cho trend analysis theo tuần. Cũng có thể filter theo Service, Linked Account, Tag, Region, Usage Type, Instance Type.
Cost Anomaly Detection có miễn phí không?

Xem đáp án

Miễn phí — Cost Anomaly Detection (2020) không có phí riêng. Sử dụng ML để detect cost anomalies so với historical patterns, gửi alert qua email hoặc SNS khi phát hiện. Có thể setup monitors theo individual service, AWS account, cost category, hoặc cost allocation tag. Phát hiện được ransomware mining, data exfiltration, code bugs gây cost spike — trước khi nhận bill cuối tháng.
Cost Allocation Tags hiện trong Cost Explorer sau bao lâu?

Xem đáp án

24 giờ sau khi activate. Điều quan trọng: chỉ cost data sau thời điểm activate mới được tag — không retroactive. Và tags phải được activate trong Billing Console (Billing → Cost Allocation Tags) — chỉ tag resource không đủ. Workflow: tag resources → activate tag → đợi 24h → filter theo tag trong Cost Explorer.
CUR khác Cost Explorer ở điểm gì?

Xem đáp án

Cost and Usage Report (CUR): raw, granular CSV/Parquet data export sang S3 — mọi line item (hourly), resource IDs, pricing details, reservation usage. Dùng với Athena, QuickSight, third-party tools cho custom analysis. Cost Explorer: interactive visualization, predefined dashboards, forecasting, recommendations — không granular như CUR nhưng dễ dùng hơn. CUR cho advanced analytics; Cost Explorer cho quick insights.
Compute Optimizer khác Cost Explorer recommendation ở điểm gì?

Xem đáp án

Compute Optimizer: dùng ML analyze CloudWatch utilization metrics (CPU, network, memory từ agent) để recommend optimal EC2 instance type, Lambda memory, EBS size — rất cụ thể với performance trade-offs. Cost Explorer right-sizing: recommend based on CloudWatch metrics đơn giản hơn, focus trên cost savings của EC2. Compute Optimizer có ML model tốt hơn, analyze Lambda và EBS — Cost Explorer chỉ EC2.

Bài tập thực hành

Mở Cost Explorer, group by Service, identify top 3 costs
Tạo monthly cost budget $50 với email alert
Enable Cost Anomaly Detection
Activate user-defined tags Project và Environment
Estimate cost cho 1 EC2 t3.medium + RDS db.t3.small via Pricing Calculator

Tài liệu tham khảo chính thức

Tiếp theo: Reserved Instances và Savings Plans Deep Dive