Billing, Cost Management, and Resource Planning
Billing foundations
Alibaba Cloud bills per resource on models: subscription (prepaid), pay-as-you-go (postpaid, metered), and spot. Costs are visible through bills, cost analysis, and budgets. Coupons, savings plans, and reserved capacity reduce steady-state cost. Resource planning means sizing to measured demand, not peak guesswork.
Cost levers
- Right-sizing - downsize over-provisioned instances; match family to workload.
- Commitment - subscription or savings plans for predictable baseline.
- Elasticity - scale in off-peak; release idle resources.
- Storage tiering - OSS lifecycle to IA/Archive; delete orphaned snapshots and unattached disks.
- Network - NAT/EIP charges accrue; audit idle EIPs and cross-zone traffic.
- Tagging and resource groups - allocate cost to teams for accountability.
Traps
- Assuming a stopped pay-as-you-go instance costs nothing - disks, EIPs, and images may still bill.
- Forgetting data transfer and NAT processing charges when estimating.
- Overlooking that snapshots and unused images accumulate silently.
- Treating spot as a guaranteed capacity source - it can be reclaimed.
- Ignoring that cross-region replication and backups add ongoing cost.
Rule of thumb
Split spend into baseline (steady, commit with subscription/savings plans) and variable (burst, use pay-as-you-go or spot). Tag everything, review monthly, and set budgets with alerts before optimizing.
Worked example
A workload runs 24/7 at 8 vCPU but peaks to 16 for 4 hours daily. Commit 8 vCPU on subscription, add 8 vCPU pay-as-you-go for the peak window, and use lifecycle rules to move cold OSS data to Archive, cutting steady-state cost without hurting peak performance.