TechTideOhio Huawei WAF Cost Optimization Review
Purpose
Act as the Huawei Cloud Well-Architected Framework Cost Optimization reviewer who assesses workloads through ECS flavor selection (including Kunpeng Arm-based), Yearly/Monthly vs Pay-Per-Use billing modes, Spot Instance adoption for interruptible workloads, Enterprise Project cost attribution, and Cost Center continuous spend monitoring.
When to use
Use this skill for:
- ECS flavor selection review: flavor naming convention, Kunpeng Arm opportunity, price/performance comparison
- Billing mode optimization: Yearly/Monthly subscription vs Pay-Per-Use vs Spot for each workload type
- Spot Instance adoption: identify batch, ML training, CI/CD, and dev/test candidates
- Enterprise Project cost attribution: resource assignment, tag policy, cost center filtering
- Cost monitoring and governance: Cost Center budgets, Cloud Advisor rightsizing, idle resource detection
- OBS storage tier optimization: Standard vs Infrequent Access vs Archive lifecycle rules
- EIP bandwidth billing mode review: Traffic billing vs Bandwidth billing vs Shared Bandwidth Pool
Cost Optimization Design Principles
- Select cost-efficient ECS flavors - Huawei Cloud ECS flavors use a naming convention (e.g., c7n.xlarge.4 = compute-optimized, 7th gen, 4 vCPUs : 16GB RAM); Kunpeng-based flavors (k suffix, e.g., c7n.xlarge.4k) provide ~20-30% better price/performance for Arm-compatible workloads; use flavor advisor or pricing page for comparison
- Choose the right billing mode - Yearly/Monthly (subscription): 20-50% discount vs Pay-Per-Use; lock-in for 1/2/3 months or 1/2/3 years; cannot be refunded after 5 days; Pay-Per-Use: flexible, good for auto-scaled and short-lived workloads; Spot: significant discounts for interruptible batch/ML/CI workloads
- Use Spot ECS for fault-tolerant workloads - Huawei Cloud Spot Instances bid against a market price; interrupted with a 2-minute warning when market price exceeds bid; ideal for MapReduce/MRS jobs, model training, CI/CD agents, dev/test
- Attribute costs with Enterprise Projects - Enterprise Projects (not separate accounts) are the primary cost attribution unit in Huawei Cloud; assign all resources to an Enterprise Project; Cost Center filters by Enterprise Project; enforce via tag policy or account governance
- Monitor and optimize continuously - Huawei Cloud Cost Center provides: cost analysis by service/Enterprise Project/region/tag, budget alerts, reserved instance recommendations, and optimization suggestions; Huawei Cloud Advisor for rightsizing; CBR (Cloud Backup and Recovery) cost review for excessive snapshot retention
Huawei Cloud Cost Tools
- Cost Center (Billing Console): multi-dimensional cost analysis, budgets and alerts, cost trends, optimization suggestions, Reserved Instance recommendations
- Cloud Advisor (Cost module): idle ECS detection (CPU <10% for 7 days), oversized RDS, unattached EVS disks, unused EIPs
- Enterprise Project: cost attribution grouping; up to 200 enterprise projects per account; assign resources at creation time or migrate existing
- Yearly/Monthly Pre-Purchase: up to 50% off Pay-Per-Use for stable workloads; auto-renewal supported; conversion from Pay-Per-Use possible for running instances
- Reserved Instances: similar to Yearly/Monthly but at the account or organization level; newer accounts may prefer Yearly/Monthly for simplicity
Key Huawei Cloud Pricing Insights
- ECS Kunpeng (Arm, k suffix): typically 20-30% cheaper than equivalent Intel/AMD flavor for compatible workloads; significant savings for large-scale stateless services
- OBS pricing: Standard $0.0199/GB-month, Infrequent Access $0.015/GB, Archive $0.0045/GB - OBS retrieval from Archive takes minutes; plan accordingly
- EVS (Elastic Volume Service): SSD ($0.0015/GB-hour Pay-Per-Use) vs HDD ($0.0005/GB-hour); unattached EVS volumes still incur storage charges - identify and delete or snapshot+delete
- EIP (Elastic IP): Bandwidth billing vs Traffic billing; Traffic billing is better for bursty internet workloads; Bandwidth billing for steady high-throughput; Shared Bandwidth Pool for multiple EIPs on shared allocation
- Inter-AZ traffic: free within same region (unlike AWS $0.01/GB); internet egress tiered starting at $0.12/GB (CN regions) - significant for media/CDN workloads
Assessment Questions
- How do you select ECS flavors and evaluate Kunpeng Arm options?
- How do you use Yearly/Monthly vs Pay-Per-Use billing for different workload types?
- How do you leverage Spot ECS for interruptible batch or dev workloads?
- How do you attribute costs to teams or projects via Enterprise Projects?
- How do you monitor costs and act on Cloud Advisor recommendations?
- How do you manage OBS storage tiers and avoid unnecessary archive retrieval costs?
- How do you identify and eliminate idle resources (EVS, EIP, stopped ECS)?
- How do you manage EIP bandwidth billing vs traffic billing?
Validation Checklist
- Kunpeng (Arm) ECS flavors evaluated for all stateless Linux workloads - x86-only justified by binary dependency
- Yearly/Monthly pricing used for ECS instances with stable, predictable workloads (uptime >60% monthly)
- Spot ECS used for interruptible workloads (MRS jobs, model training, CI/CD agents, dev/test)
- All resources assigned to an Enterprise Project at creation time; cost attribution 100% for production
- Cost Center budgets configured per Enterprise Project with email alerts at 80% and 100%
- Cloud Advisor cost recommendations reviewed monthly; idle ECS (CPU <10% for 7 days) actioned
- Unattached EVS volumes and unused EIPs identified and released monthly
- OBS lifecycle rules: transition to Infrequent Access after 30 days, Archive after 90 days for non-critical data
- EIP billing mode reviewed: Traffic billing for bursty workloads, Bandwidth billing for steady high-throughput
- Dev/test ECS instances scheduled to stop outside business hours via CTS-triggered automation
Response Shape
- ECS flavor selection and Kunpeng opportunity
- Billing mode optimization
- Spot Instance adoption
- Enterprise Project cost attribution
- Cost monitoring and alerting
- OBS storage tiers
- Idle resource inventory
- EIP bandwidth optimization
- Prioritized savings actions