Reading25 min read·Module 4High exam weight

Savings Plans

Key concepts

  • Compute Savings Plans

  • EC2 Instance Savings Plans

  • SageMaker Savings Plans

  • Flexibility vs discount

  • Savings Plans vs Reserved

Overview

AWS Savings Plans offer a flexible pricing model with significant discounts (up to 72%) in exchange for a commitment to a consistent amount of usage ($/hour) for 1 or 3 years. Unlike Reserved Instances, Savings Plans are based on dollar commitment rather than specific instance configurations.

Core Concept

Savings Plans work on a dollar commitment (e.g., $10/hour for 1 year) rather than instance commitment. Compute Savings Plans offer maximum flexibility (any region, any instance, plus Lambda and Fargate). EC2 Instance Savings Plans offer higher discounts but lock you to an instance family and region.

Exam Tip

When exam questions mention 'flexibility to change instance families, regions, or use Lambda/Fargate' while still getting discounts, Compute Savings Plans is the answer. When the scenario mentions 'committed to a specific instance family' but wants flexibility in size/OS, EC2 Instance Savings Plans applies.

Key Concepts

Types of Savings Plans

AWS Savings Plans Types Comparison
Figure 1: Savings Plans Types and Coverage

Compute Savings Plans

Coverage: EC2, Fargate, and Lambda (any region, any configuration)

Maximum Discount: Up to 66%

Flexibility:

  • Any EC2 instance family, size, OS, tenancy
  • Any AWS region
  • Applies to Fargate and Lambda usage
  • Automatically applies to most cost-effective usage first

Key Features:

  • Most flexible savings option
  • No need to specify instance details
  • Ideal for workloads that may change or migrate
  • Covers containerized and serverless workloads

Best For:

  • Organizations with evolving infrastructure
  • Multi-region deployments
  • Mixed EC2 + Lambda + Fargate workloads
  • Companies planning cloud modernization

EC2 Instance Savings Plans

Coverage: EC2 only, specific instance family and region

Maximum Discount: Up to 72%

Flexibility:

  • Any instance SIZE within the family
  • Any operating system
  • Any tenancy (shared, dedicated, host)
  • Single region commitment

Commitment Required:

  • Instance family (e.g., m5, c5, r5)
  • Region (e.g., us-east-1)
  • NOT locked to: size, OS, or tenancy

Best For:

  • Stable workloads with known instance family needs
  • Single-region deployments
  • Maximum savings with moderate flexibility
  • Organizations with predictable compute patterns

SageMaker Savings Plans

Coverage: Amazon SageMaker ML instances

Maximum Discount: Up to 64%

Applies To:

  • SageMaker Studio Notebooks
  • SageMaker On-Demand Notebooks
  • SageMaker Processing
  • SageMaker Data Wrangler
  • SageMaker Training
  • SageMaker Real-Time Inference
  • SageMaker Batch Transform

Flexibility:

  • Any instance family, size, or region
  • Automatically applies to eligible SageMaker usage

Best For:

  • Organizations with significant ML workloads
  • Teams using SageMaker for production ML

Savings Plans vs Reserved Instances

Savings Plans vs Reserved Instances Comparison
Figure 2: Flexibility vs Discount Comparison

Savings Plans vs Reserved Instances

FeatureCompute SPEC2 Instance SPStandard RIConvertible RI
Max Discount66%72%72%66%
Commitment Type$/hour$/hourInstancesInstances
Region FlexibilityAnyFixedFixedFixed (exchange)
Instance Family FlexAnyFixedFixedExchange only
Instance Size FlexAnyAnyModify onlyExchange only
OS FlexibilityAnyAnyFixedExchange only
Lambda/FargateYesNoNoNo
Sell on MarketplaceNoNoYesNo
Capacity ReservationNoNoYes (Zonal)Yes (Zonal)

When to Choose Each

Choose Compute Savings Plans When:

  • Workloads span EC2, Lambda, and Fargate
  • You may change instance families
  • You may migrate between regions
  • Maximum flexibility is priority

Choose EC2 Instance Savings Plans When:

  • Workload is stable on known instance family
  • Single region deployment
  • Want higher discount than Compute SP
  • Only need EC2 (no Lambda/Fargate)

Choose Standard Reserved Instances When:

  • Need capacity reservation guarantee
  • Want to resell on RI Marketplace
  • Highest discount is priority
  • Configuration won't change

Choose Convertible Reserved Instances When:

  • Need capacity reservation
  • May need to exchange configurations
  • Configuration is likely to evolve

How Savings Plans Apply

Automatic Application

Application Order:

  1. Standard Reserved Instances (capacity reservation first)
  2. Convertible Reserved Instances
  3. Savings Plans (most expensive On-Demand first)
  4. Spot Instances (if applicable)
  5. On-Demand (remainder)

Optimization Logic:

  • Savings Plans automatically apply to usage that provides maximum savings
  • Most expensive eligible On-Demand usage covered first
  • Updates hourly as usage changes

Example:

  • Commitment: $10/hour Compute Savings Plan
  • Running: c5.xlarge ($0.17/hr), m5.large ($0.096/hr), Lambda
  • SP applies to c5.xlarge first (higher rate = more savings)
  • Remaining commitment covers other usage
TEXTCalculate Savings Plan Commitment
Scenario: Steady usage of 10x m5.large (Linux) running 24/7

Step 1: Calculate On-Demand hourly cost
  - m5.large On-Demand: $0.096/hour
  - 10 instances: 10 × $0.096 = $0.96/hour

Step 2: Choose Savings Plan type
  - Compute SP discount: ~43% → $0.055/hour per instance
  - EC2 Instance SP discount: ~53% → $0.045/hour per instance

Step 3: Calculate commitment
  - Compute SP: 10 × $0.055 = $0.55/hour commitment
  - EC2 Instance SP: 10 × $0.045 = $0.45/hour commitment

Step 4: Monthly/Yearly cost
  - Compute SP: $0.55 × 730 hours = $401.50/month
  - EC2 Instance SP: $0.45 × 730 hours = $328.50/month
  - On-Demand: $0.96 × 730 hours = $700.80/month

Annual Savings:
  - Compute SP: $3,592/year savings (43%)
  - EC2 Instance SP: $4,467/year savings (53%)
SHAWS CLI - Savings Plans Commands
# List Savings Plans
aws savingsplans list-savings-plans

# Describe specific Savings Plan
aws savingsplans describe-savings-plans \
  --savings-plan-ids sp-1234567890abcdef0

# Get Savings Plans utilization
aws ce get-savings-plans-utilization \
  --time-period Start=2024-01-01,End=2024-01-31

# Get Savings Plans coverage
aws ce get-savings-plans-coverage \
  --time-period Start=2024-01-01,End=2024-01-31

# Get Savings Plans purchase recommendation
aws ce get-savings-plans-purchase-recommendation \
  --savings-plans-type COMPUTE_SP \
  --term-in-years ONE_YEAR \
  --payment-option NO_UPFRONT \
  --lookback-period-in-days SIXTY_DAYS

Payment Options

Savings Plans Payment Options

Payment OptionDiscount LevelCash FlowBest For
All UpfrontHighestPay full amount nowMaximum savings, available capital
Partial UpfrontMediumSome now, rest monthlyBalance of savings and cash flow
No UpfrontLowestMonthly payments onlyPreserve cash, still get savings

Term Options

1-Year vs 3-Year Terms

1-Year Term

  • Lower commitment risk
  • Smaller discount (but still significant)
  • Easier to adapt to changing needs
  • Good for evolving workloads

3-Year Term

  • Maximum discount potential
  • Longer commitment period
  • Best for stable, long-term workloads
  • Higher risk if needs change

Recommendation:

  • New workloads: Start with 1-year
  • Proven stable workloads: Consider 3-year
  • Uncertain growth: 1-year with review

Best Practices

  1. Start with Cost Explorer: Analyze usage patterns for 30+ days before purchasing
  2. Use Recommendations: AWS provides purchase recommendations based on historical usage
  3. Cover Baseline First: Size Savings Plans for consistent minimum usage, not peaks
  4. Combine Strategies: Use Savings Plans for steady-state, Spot/On-Demand for variable
  5. Monitor Utilization: Regularly check Savings Plans utilization in Cost Explorer
  6. Choose Based on Flexibility Needs: Don't over-commit if changes are likely
  7. Consider All Compute: Compute SP covers Lambda/Fargate; don't forget these

Common Exam Scenarios

Exam Scenario Decision Guide

ScenarioRecommended SolutionKey Reasoning
Multi-region EC2 deployment, may migrate regionsCompute Savings PlansRegion flexibility needed
Steady m5 workload in us-east-1, size may changeEC2 Instance Savings PlansHigher discount, locked to family/region is OK
EC2 now, planning Lambda migrationCompute Savings PlansCovers both EC2 and Lambda
Need capacity guarantee in specific AZReserved Instances (Zonal)Only RIs provide capacity reservation
SageMaker training jobs varying in sizeSageMaker Savings PlansSpecific to SageMaker workloads
Want to sell unused capacity laterStandard Reserved InstancesOnly Standard RIs can be sold
Mixed Fargate + Lambda + EC2 workloadCompute Savings PlansCovers all three services
Maximum savings, workload very stableEC2 Instance SP (3-year, All Upfront)Highest discount for stable workloads

Common Pitfalls

Over-Commitment

Savings Plans charge for the committed amount whether you use it or not. Over-committing based on peak usage rather than baseline results in paying for unused commitment. Always size for minimum steady-state usage.

Wrong Plan Type Selection

Choosing EC2 Instance Savings Plans and then migrating to Lambda or Fargate means the Savings Plan doesn't apply. If you're considering containerization or serverless, choose Compute Savings Plans even though the discount is slightly lower.

Cannot Be Cancelled or Sold

Unlike Standard Reserved Instances, Savings Plans CANNOT be sold on a marketplace. Once purchased, you're committed for the full term. Plan carefully before purchasing.

No Capacity Reservation

Savings Plans do NOT provide capacity reservations. If you need guaranteed capacity in a specific AZ (e.g., for disaster recovery), you still need Reserved Instances or On-Demand Capacity Reservations.

Ignoring Lambda/Fargate Coverage

Organizations often purchase EC2 Instance Savings Plans without considering their Lambda or Fargate usage. Compute Savings Plans cover all three, potentially providing better overall value.

Quick Reference

Savings Plans Comparison

Plan TypeMax DiscountApplies ToFlexibility
Compute SP66%EC2, Lambda, FargateAny region/family
EC2 Instance SP72%EC2 onlyFixed family/region
SageMaker SP64%SageMakerAny region/family

Key Terms

TermDescription
Commitment$/hour you commit to for the term
Utilization% of commitment actually used
Coverage% of eligible usage covered by SP
Term1 year or 3 years

CLI Quick Reference

# Get purchase recommendations
aws ce get-savings-plans-purchase-recommendation \
  --savings-plans-type COMPUTE_SP \
  --term-in-years ONE_YEAR \
  --payment-option NO_UPFRONT \
  --lookback-period-in-days THIRTY_DAYS

# Check utilization
aws ce get-savings-plans-utilization \
  --time-period Start=2024-01-01,End=2024-01-31 \
  --granularity MONTHLY

# List active Savings Plans
aws savingsplans list-savings-plans --states active

Test Your Knowledge

Q

A company runs EC2 workloads across multiple regions and is considering migrating some workloads to Lambda in the next year. Which pricing option provides BOTH savings and the required flexibility?

AStandard Reserved Instances
BEC2 Instance Savings Plans
CCompute Savings Plans
DConvertible Reserved Instances
Q

A company has stable production workloads running exclusively on m5 instances in us-east-1. They want MAXIMUM cost savings and don't plan to change instance families or regions. Which option provides the highest discount?

ACompute Savings Plans (3-year, All Upfront)
BEC2 Instance Savings Plans (3-year, All Upfront)
CStandard Reserved Instances (3-year, All Upfront)
DConvertible Reserved Instances (3-year, All Upfront)
Q

A company purchased EC2 Instance Savings Plans for $5/hour but their usage has dropped and they're only using $3/hour worth of compute. What happens to the unused $2/hour?

Q

A startup needs guaranteed EC2 capacity in a specific Availability Zone for their disaster recovery setup. Which option provides BOTH cost savings AND capacity reservation?

ACompute Savings Plans with On-Demand Capacity Reservation
BEC2 Instance Savings Plans
CZonal Reserved Instances
DRegional Reserved Instances

Further Reading

Related services

Savings Plans