Skip to content

Purchase-Option Cloud-Cost Optimization

1 · Purpose

Maximize long-term AWS savings by recommending the optimal mix of Savings Plans (SPs), Reserved Instances (RIs), and Spot—across all linked accounts and services—while making trade-offs transparent to finance and engineering.

2 · Capabilities

Checklist Item ZirconTech Implementation
1- & 3-year SP/RI recommendations Athena CUR analysis + AWS Compute Optimizer SP/RDS RI CSV exports; forecasts normalized usage to recommend commitments for both 1- and 3-year terms
Cross-account coverage Pull CUR “LinkedAccountId” + Organizations tag to aggregate usage; recommendations split by payer or BU
All SP/RI types Compute SP, EC2 Instance SP, Standard/Convertible, Zonal/Regional RIs; coverage plan JSON lists each recommendation
Multi-service RIs Same pipeline evaluates RDS, Redshift, ElastiCache, DynamoDB standby; uses respective service recommendation APIs
Spot instance guidance Spot Fleet Advisor + AWS Instance Selector CLI to build diversified launch templates matched to workload tolerance

3 · Methodology

3.1 Data Collection

  1. Enable Cost & Usage Report with hourly granularity.
  2. Run aws compute-optimizer export-recommendation for EC2/RDS.
  3. Ingest to Athena table finops.rec_input.

3.2 Commitment Modeling

Step Tool Output
Calculate normalized hours Athena SQL vw_norm_usage
Run RI/SP optimizer Python notebook w/ boto3 + OR-Tools JSON plan (ri_sp_plan.json)
Spot workload fit Instance Selector + historical price API “Spot-eligible” launch template list

3.3 Decision Workshop

  • Compare options: Compute SP vs. EC2 Instance SP, SP vs. RI, visualized in QuickSight “Commitment Trade-off” dashboard.
  • Finance chooses risk appetite (commit %).
  • CAB approves purchase order or Private Offer.

3.4 Execution & Tracking

  • Commit via AWS Console or purchase-savings-plan API.
  • Tag commitments CostCenter, Project.
  • Monthly variance check: actual vs. covered hours.

4 · Clarify / Compare / Contrast

Feature Compute SP EC2 Instance SP Standard RI Convertible RI
Flexibility Any region, OS, size Specific family & region Exact family/size/AZ Family swap allowed
Discount 66 % max 72 % max Up to 75 % Up to 54 %
Applies to EC2, Fargate, Lambda EC2 only EC2 only EC2 only
Convertible? N/A N/A No Yes
Best for Diverse workloads Homogeneous fleets Steady state in one AZ Long-lived but changing

5 · Spot Instance Recommendations

Workload Recommendation Rationale
Batch / CI jobs Spot Fleet with 6 instance pools spread across 3 AZs 90 % savings, interruption-tolerant
Stateless web EC2 Auto Scaling group: 20 % Spot, 80 % On-Demand Balances savings and availability
Stateful DB Do not use Spot Data loss risk

Instance-type optimization

aws ec2-instance-selector --vcpus 4 --memory 16 --usage-class spot \
  --price-per-hour 0.05 --region us-east-1
````

Outputs cheapest compatible types considering real-time Spot price.

---

## 6 · Sample Coverage Plan (inline)

```json
{
  "generated": "2025-06-30T12:00:00Z",
  "computeSavingsPlan": {
    "termYears": 3,
    "hourlyCommitUSD": 5.2,
    "expectedAnnualSavingsUSD": 21_300
  },
  "ec2InstanceSavingsPlan": {
    "m6i.large": { "commitHours": 8, "savings": 2_800 }
  },
  "rdsStandardRI": {
    "db.r7g.large": { "term": "1yr", "paymentOption": "NoUpfront", "count": 2 }
  },
  "spotStrategy": { "fleetDiversification": 6, "targetPct": 25 }
}

7 · Continuous Optimization Loop

Frequency Task Owner
Weekly Import latest usage, refresh commitment model Lambda
Monthly Governance call to review coverage & Spot savings FinOps Analyst
Qtrly Rebalance Convertible RI exchanges Solutions Architect

Last updated: 30 Jun 2025