SkarpChapter 16 of 20
AWS Pricing Models and Amazon EC2 Cost Strategies
Turn pricing tables into practical decisions by learning how AWS pricing models work and how EC2 purchasing options affect your bill.
Pricing Mindset: How AWS Charges for Cloud
Why Pricing Models Matter
For the Cloud Practitioner exam you do not memorize prices. You must understand how AWS charges and how to choose between pricing models and EC2 purchasing options.
Three Big Ideas
AWS pricing is built around: 1) pay only for what you use, 2) discounts when you commit, and 3) discounts when you are flexible about when and how you run workloads.
What Drives EC2 Cost
Your EC2 bill depends on: instance type/size, Region, how long instances run, and especially which purchasing option you choose (On-Demand, Reserved, Savings Plans, Spot).
What You Will Practice
You will compare pricing models, relate them to workload patterns (steady, spiky, fault-tolerant), and answer scenario-style questions similar to the certification exam.
Core AWS Pricing Models: Pay-as-you-go and Commitments
Pay-as-you-go
Pay-as-you-go means you pay only for what you use, with no long-term commitment. On-Demand EC2, S3 storage, and Lambda invocations are classic examples.
Commitment-based Discounts
With Reserved Instances and Savings Plans, you commit to consistent usage (typically 1 or 3 years) in exchange for a lower hourly or per-unit rate.
Payment and Duration Options
Commitments usually offer 1-year or 3-year terms and payment options: All Upfront, Partial Upfront, or No Upfront. More upfront often means a bigger discount.
Volume and Tiered Pricing
Many services use volume or tiered pricing: as your usage grows, the price per unit decreases. S3 and data transfer are common examples.
Exam Mapping
Map concepts: On-Demand = flexibility, Reserved/Savings Plans = steady usage savings, Spot = very cheap but interruptible.
Amazon EC2 On-Demand, Reserved, and Spot: The Big Picture
Three EC2 Pricing Options
For EC2, you must know: On-Demand, Reserved Instances (RIs), and Spot Instances. Each trades off cost versus flexibility and risk.
On-Demand Instances
On-Demand has no long-term commitment, billed per second or hour. It costs more per unit but gives maximum flexibility for new, short, or spiky workloads.
Reserved Instances
Reserved Instances require a 1- or 3-year commitment to specific instance attributes, in exchange for a significant discount. Standard RIs give more savings; Convertible RIs give more flexibility.
Spot Instances
Spot Instances use spare EC2 capacity at very low prices, but AWS can interrupt them. They are ideal for fault-tolerant, flexible workloads like batch jobs and CI/CD.
Cost vs Risk Trade-off
As you move from On-Demand → RIs → Spot, price decreases but risk and constraints increase. Choosing correctly depends on workload behavior.
Savings Plans vs Reserved Instances (Current Best Practice)
What Are Savings Plans?
Savings Plans are a flexible, commitment-based discount model. You commit to spend a certain USD/hour on compute for 1 or 3 years, and AWS discounts that usage.
Types of Savings Plans
Compute Savings Plans apply broadly across EC2, Fargate, and Lambda and Regions. EC2 Instance Savings Plans focus on one instance family in one Region for higher discounts.
Savings Plans vs RIs
Standard RIs can give slightly higher discounts but are rigid. Convertible RIs are more flexible but need manual exchanges. Savings Plans adjust automatically within their scope.
Exam Clue: Multi-service Compute
If a scenario mentions simplifying discounts across EC2, Fargate, and Lambda, Savings Plans are the best-fit answer, not traditional RIs.
Current Best Practice
As of 2026, new compute commitments usually use Savings Plans. RIs still exist and appear on the exam, especially in EC2-specific cost questions.
Workload-Based Decisions: Steady, Spiky, and Fault-Tolerant
Steady-State Workloads
Always-on or predictable workloads (like production databases) are ideal for Savings Plans or Reserved Instances covering the baseline, with small On-Demand for variations.
Spiky or Unpredictable Workloads
For spiky traffic, use On-Demand for sudden peaks. You can cover a known minimum with commitments and let Auto Scaling add On-Demand instances as needed.
Fault-Tolerant Workloads
If jobs can be interrupted and restarted, use Spot Instances heavily. This suits batch jobs, analytics, and CI/CD where interruptions are acceptable.
Real Example: E-commerce
An online store might commit to a baseline of 4 instances (Savings Plans) and let Auto Scaling add On-Demand or Spot capacity during flash sales.
Exam Pattern Recognition
When you see steady, think commitments. When you see unpredictable, think On-Demand. When you see fault-tolerant batch, think Spot.
Thought Exercise: Mix and Match EC2 Options
Use this step to reason like a cost-aware cloud practitioner. For each scenario, pause and decide which EC2 pricing options you would combine. Then compare with the suggested approach.
- Startup API Backend
- Scenario: A new startup launches an API. They have no real traffic history. They expect to change instance types as they learn more.
- Your choice: Which EC2 purchasing option is safest at the beginning?
- Suggested answer: On-Demand Instances. They avoid lock-in while the architecture and usage are still uncertain. After a few months of stable usage, they can add Savings Plans.
- Long-running Internal Tool
- Scenario: An internal HR system runs on EC2 and has been stable for over a year. It must be available during business hours, and usage is predictable.
- Your choice: Which model offers the best long-term savings?
- Suggested answer: Use Savings Plans or RIs for the predictable baseline. If the system only runs 9–5, they might also consider scheduling instances to shut down outside business hours to avoid paying when idle.
- University Research Batch Jobs
- Scenario: A research group runs large simulations at night on EC2. Jobs can be re-run if interrupted.
- Your choice: How do you minimize cost while keeping reasonable throughput?
- Suggested answer: Use a mix of Spot Instances (for most of the capacity) plus a small number of On-Demand instances to guarantee some progress even when Spot capacity is limited.
As you go through later exam-style questions in Skarp, practice this same reasoning: identify the workload pattern first, then map to pricing options.
Other Cost Factors: Storage, Data Transfer, and Free Tier
Storage and EC2
EC2 commonly uses EBS volumes. You pay for provisioned GBs, some performance options, and backups as S3 snapshots. Storage choices affect total cost.
S3 Storage Classes
S3 offers classes like Standard, Infrequent Access, and Glacier. Lower storage price often means trade-offs such as higher retrieval costs or slower access.
Data Transfer Basics
Data into AWS is usually free; data out to the internet is charged. Cross-AZ and cross-Region traffic also has costs, so architecture impacts network charges.
Free Tier Overview
The AWS Free Tier gives new accounts limited free usage, plus some always-free offers. It is not a pricing model but helps customers experiment cheaply.
Holistic Cost View
For real workloads and exam scenarios, consider compute + storage + data transfer, not just EC2 hourly rates, when optimizing cost.
Quick Check: Mapping Workloads to EC2 Options
Test your ability to pick the right EC2 pricing model from a short scenario.
A company has a batch data processing job that runs every night. The job can be safely interrupted and restarted without affecting customers. They want to minimize compute cost. Which EC2 pricing option is the BEST primary choice?
- On-Demand Instances only
- Reserved Instances for all capacity
- Spot Instances for most capacity, possibly mixed with some On-Demand
- Dedicated Hosts for all instances
Show Answer
Answer: C) Spot Instances for most capacity, possibly mixed with some On-Demand
Fault-tolerant, batch-style workloads that can be interrupted are ideal for **Spot Instances**, which offer very low prices but may be reclaimed. Many real-world designs mix Spot (for most capacity) with a small On-Demand baseline. On-Demand only is more expensive, Reserved Instances are meant for steady, always-on usage, and Dedicated Hosts are for licensing/compliance, not cost optimization.
Quick Check: Savings Plans vs RIs
Check your understanding of modern commitment-based discounts for compute.
An organization runs a mix of EC2 instances, AWS Fargate tasks, and AWS Lambda functions. They want a single commitment model to reduce overall compute costs with minimal management overhead. Which option is the BEST fit?
- Standard Reserved Instances for all EC2 instances
- Compute Savings Plans
- EC2 Instance Savings Plans for each instance family
- Spot Instances for all workloads
Show Answer
Answer: B) Compute Savings Plans
Only **Compute Savings Plans** apply across EC2, Fargate, and Lambda and automatically cover eligible usage up to the committed spend. EC2 Instance Savings Plans and RIs are limited to EC2 and sometimes specific families/Regions. Spot Instances are not a commitment model and cannot be used for all workloads (for example, not ideal for all Lambda or critical production services).
Key Terms: AWS Pricing and EC2 Options
Use these flashcards to reinforce core pricing vocabulary before moving on.
- On-Demand Instances
- EC2 instances with no long-term commitment. You pay for compute capacity by the second or hour, with the highest price per unit but maximum flexibility. Ideal for new, short, or unpredictable workloads.
- Reserved Instances (RIs)
- A commitment-based discount for EC2 where you commit to a specific instance configuration in a Region for 1 or 3 years. Standard RIs offer higher discounts with less flexibility; Convertible RIs offer more flexibility with somewhat lower discounts.
- Spot Instances
- EC2 instances that use spare AWS capacity at steep discounts. They can be interrupted by AWS with short notice, so they are best for fault-tolerant, flexible workloads like batch processing and CI/CD.
- Savings Plans
- A flexible pricing model where you commit to a specific amount of compute spend per hour (for example, 10 USD/hour) for 1 or 3 years, in exchange for discounted rates on eligible usage.
- Compute Savings Plans
- The most flexible Savings Plan type. Applies to EC2, AWS Fargate, and AWS Lambda across instance families, operating systems, and Regions, up to the committed spend.
- EC2 Instance Savings Plans
- A Savings Plan type that applies to a specific EC2 instance family in a chosen Region. Less flexible than Compute Savings Plans but offers higher discounts.
- Steady-state workload
- A workload with predictable, long-running usage (often 24/7). Typically optimized with commitment-based discounts like Reserved Instances or Savings Plans.
- Fault-tolerant workload
- A workload that can handle interruptions, delays, or restarts without major impact. Well-suited to Spot Instances for significant cost savings.
- Volume-based pricing
- A pricing pattern where the cost per unit decreases as usage increases, commonly seen in services like Amazon S3 and data transfer.
- AWS Free Tier
- A set of limited, free usage offers for new AWS customers (and some always-free services) that allows experimentation without significant cost. It is not a separate pricing model.
Putting It Together: A Simple Cost Optimization Plan
Scenario Overview
A learning platform has 4 always-on web EC2 instances, up to 8 extra during peaks, a nightly analytics job, data in S3, and RDS. Usage is stable and expected to continue.
Baseline Commitments
The 4 always-on web instances are steady-state. Cover them with Savings Plans or Reserved Instances to reduce long-term cost.
Handling Peaks
The extra 0–8 instances are spiky. Keep them as On-Demand, or use a mix of On-Demand and Spot in an Auto Scaling group if the app tier is stateless.
Batch Analytics on Spot
The nightly analytics job is fault-tolerant, so it is ideal for Spot Instances. This can greatly reduce compute cost for that workload.
Beyond EC2 and Commitments
Optimize S3 storage classes and data transfer. Because they may adopt Fargate, Compute Savings Plans give flexible discounts across future compute choices.
Key Terms
- AWS Free Tier
- A collection of limited free usage offers for new AWS customers, plus some always-free services, allowing low-cost experimentation.
- Savings Plans
- A flexible discount model where you commit to a specific amount of compute spend per hour for 1 or 3 years, receiving discounted rates on eligible usage.
- Spot Instances
- EC2 instances that use spare AWS capacity at steep discounts but can be interrupted by AWS with short notice, best for fault-tolerant workloads.
- Reserved Instances
- A commitment-based discount for EC2 where you commit to a specific instance configuration in a Region for 1 or 3 years, in exchange for a lower rate than On-Demand.
- On-Demand Instances
- EC2 instances with no long-term commitment. You pay for compute capacity by the second or hour, with the highest price per unit but maximum flexibility.
- Volume-based pricing
- A pricing pattern where the unit cost decreases as usage increases, used by services like Amazon S3 and data transfer.
- Compute Savings Plans
- Savings Plans that apply to EC2, AWS Fargate, and AWS Lambda across instance families, operating systems, and Regions.
- Steady-state workload
- A workload with predictable, long-running usage, often suited to commitment-based discounts like Reserved Instances or Savings Plans.
- Fault-tolerant workload
- A workload that can handle interruptions, delays, or restarts without major impact, making it a good candidate for Spot Instances.
- EC2 Instance Savings Plans
- Savings Plans that apply to a specific EC2 instance family in a chosen Region, offering higher discounts with less flexibility.