Reserved Instances & Savings Plans: Save 40-70%

Reserved Instances and Savings Plans are commitment-based pricing models that save 40-70% on compute costs across AWS, Azure, and GCP in exchange for a 1- or 3-year commitment — the single biggest lever for cloud cost reduction.

What You’ll Learn

• AWS Savings Plans (Compute, EC2, and SageMaker)
• AWS Reserved Instances (Standard vs Convertible)
• Azure Reserved VM Instances and Savings Plans
• GCP Committed Use Discounts
• Term lengths (1-year vs 3-year)
• Payment options (All Upfront, Partial Upfront, No Upfront)
• Regional vs zonal reservations
• Selling unused reservations on the marketplace

Why It Matters

On-demand pricing is the most expensive way to run cloud workloads. For any steady-state workload — databases, production web servers, Kubernetes worker nodes — reserved capacity saves 40-70%. DodaTech saves $18k/year on Durga Antivirus Pro’s cloud infrastructure by combining 3-year Compute Savings Plans with Convertible RIs.

    flowchart LR
    A[Workload Analysis] --> B[Steady State?]
    B -->|Yes| C[Savings Plans]
    B -->|Yes| D[Reserved Instances]
    B -->|Yes| E[Committed Use Discounts]
    C --> F[Compute SP: 30-66%]
    D --> G[Standard RI: 30-60%]
    E --> H[GCP CUD: up to 70%]
    B -->|No| I[Spot / On-Demand]
    F --> J[40-70% Total Savings]
    style J fill:#f59e0b,color:#fff
  

1. AWS Savings Plans

AWS Savings Plans are flexible pricing models that apply to any eligible compute usage within a plan category.

Compute Savings Plan

Applies to EC2, Lambda, and Fargate across any region, any instance family, any OS, and any tenancy.

# Purchase a Compute Savings Plan
aws savingsplans create-savings-plan \
  --savings-plan-offering-id sp-offering-123 \
  --commitment 100.00 \
  --term 1year \
  --payment-option PartialUpfront

# Check savings plan utilization
aws savingsplans describe-savings-plans \
  --savings-plan-arns arn:aws:savingsplans:us-east-1:123456789012:savingsplan/sp-123

EC2 Instance Savings Plan

Higher discount but locked to an instance family within a region.

2. AWS Reserved Instances

Reserved Instances are tied to a specific instance type in a specific Availability Zone.

# Purchase a Standard RI
aws ec2 purchase-reserved-instances-offering \
  --instance-count 3 \
  --reserved-instances-offering-id ri-offering-123 \
  --limit-price 0.50

# List active RIs
aws ec2 describe-reserved-instances \
  --filters "Name=state,Values=active" \
  --query "ReservedInstances[].{Type:InstanceType,Count:InstanceCount,End:End}"

Standard vs Convertible:

3. Azure Reserved VM Instances

Azure RIs apply to specific VM sizes in a region. They include the Azure Savings Plan for more flexibility.

# Purchase Azure Reserved VM Instance
az reservation purchase \
  --reservation-order-id "order-456" \
  --applied-scope-type Shared \
  --sku Standard_D4s_v3 \
  --location eastus \
  --quantity 5 \
  --term P3Y \
  --billing-plan Upfront

# List reservations and savings
az reservation list \
  --query "[].{Name:name, SKU:sku, Savings:properties.savings}" \
  --output table

Azure Hybrid Benefit stacking: Combine RIs with Azure Hybrid Benefit for maximum savings — RI on compute + no license cost for Windows/SQL.

4. GCP Committed Use Discounts

GCP CUDs are purchased at the project level and apply to vCPU, memory, GPU, and even BigQuery slots.

# Purchase a GCP committed-use discount
gcloud compute commitments create prod-cud \
  --region us-central1 \
  --plan 36-month \
  --type vcpu-memory \
  --resources vcpu=100,memory=400GB

# List commitments
gcloud compute commitments list --region us-central1

CUD pricing example (e2-standard-8):

5. Term Lengths and Payment Options

Term lengths

Payment options

AWS offers three payment options that affect the effective discount:

# payment_option_comparison.py
def calculate_effective_discount(on_demand_hourly, payment_option, term_years):
    hours = term_years * 365 * 24
    on_demand_total = on_demand_hourly * hours
    
    multipliers = {
        "No Upfront": 0.65,      # 35% discount
        "Partial Upfront": 0.55, # 45% discount  
        "All Upfront": 0.50,     # 50% discount
    }
    
    total = on_demand_total * multipliers[payment_option]
    rate = total / hours
    savings = (1 - multipliers[payment_option]) * 100
    
    return {
        "rate": f"${rate:.4f}/hr",
        "total": f"${total:,.0f}",
        "savings": f"{savings:.0f}%"
    }

for option in ["No Upfront", "Partial Upfront", "All Upfront"]:
    result = calculate_effective_discount(0.384, option, 3)
    print(f"{option:<20} Rate: {result['rate']:<12} Total: {result['total']:<10} Savings: {result['savings']}")

Expected output:

No Upfront           Rate: $0.2496/hr    Total: $6,561     Savings: 35%
Partial Upfront      Rate: $0.2112/hr    Total: $5,553     Savings: 45%
All Upfront          Rate: $0.1920/hr    Total: $5,048     Savings: 50%

6. Regional vs Zonal

# AWS: Purchase zonal RI for capacity reservation
aws ec2 purchase-reserved-instances-offering \
  --instance-count 2 \
  --reserved-instances-offering-id zonal-ri-123 \
  --availability-zone us-east-1a

Recommendation: Use regional RIs for most workloads. Use zonal RIs when you need guaranteed capacity in a specific AZ (e.g., for latency-sensitive workloads colocated with other services).

7. Selling Unused Reservations

AWS allows selling unused Standard RIs on the Reserved Instance Marketplace. Convertible RIs and Savings Plans cannot be sold.

# List RIs available for sale
aws ec2 describe-reserved-instances-listings \
  --reserved-instances-id ri-12345678

# Sell a reservation
aws ec2 create-reserved-instances-listings \
  --reserved-instances-id ri-12345678 \
  --instance-count 2 \
  --pricing-schedules '[{"Term": 12, "Price": 0.15}]'

Common Mistakes

1. Over-committing without forecasting: Buying 3-year All Upfront for variable workloads leads to wasted spend. Start with 1-year Partial Upfront until usage stabilizes.

2. Ignoring Savings Plans: RIs lock you to specific instance types. Compute Savings Plans offer similar discounts with full flexibility.

3. No Hybrid Benefit stacking (Azure): Buying RIs without enabling Azure Hybrid Benefit means you pay both compute AND license costs unnecessarily.

4. Not monitoring utilization: Unused reservations are wasted money. Track utilization and sell or exchange underutilized RIs.

5. Paying All Upfront without cash flow analysis: All Upfront offers the highest discount but requires capital. Startups may prefer No Upfront to preserve cash.

Practice Questions

1. What is the difference between a Compute Savings Plan and an EC2 Instance Savings Plan? Answer: Compute SP applies to EC2, Lambda, and Fargate across any region/family. EC2 Instance SP applies to a specific instance family in a region but offers higher discounts.

2. When should you choose 3-year vs 1-year commitment? Answer: Choose 3-year for truly steady-state workloads (databases, production servers). Choose 1-year for workloads that may change or grow unpredictably.

3. Can you sell unused AWS Reserved Instances? Answer: Yes, Standard RIs can be sold on the Reserved Instance Marketplace. Convertible RIs and Savings Plans cannot be sold.

4. How do Azure RIs and Azure Savings Plans differ? Answer: Azure RIs apply to specific VM sizes; Azure Savings Plans apply to any compute service across any region, similar to AWS Compute Savings Plans.

Challenge

Design a commitment strategy for a $120k/month multi-cloud infrastructure: analyze 6 months of usage to determine baseline compute, purchase Compute Savings Plans for 70% of AWS baseline, Azure RIs for 80% of SQL VMs, GCP 3-year CUDs for BigQuery slots, set up utilization monitoring dashboards, and create a quarterly review process to adjust commitments.

FAQ

What’s Next

Related topics: Cloud Cost Optimization, Cloud Computing, FinOps

Built by the developers of Doda Browser, DodaZIP, and Durga Antivirus Pro.