AWS SAA · Domain 4 of 4

AWS SAA Domain 4: Design Cost-Optimized Architectures

This domain covers designing for cost efficiency: choosing cost-effective storage, compute, database and network options, and using pricing models, tiering and right-sizing to spend only on what the workload needs.

Exam format
20%
Official task statements
4
Assessment estimate
~10
Suggested study time estimate
20–30 hrs

About 11 min read Reviewed October 2026

Overview

Domain 4 covers cost optimization across four task statements: cost-optimized storage, compute, databases and networking. It covers S3 storage classes, lifecycle policies and Intelligent-Tiering, EC2 pricing models (On-Demand, Reserved, Savings Plans, Spot), right-sizing, serverless cost models, database cost choices, and network cost factors such as data transfer, NAT gateways and VPC endpoints.

At 20%, this is the smallest domain, but cost questions appear throughout the exam because every design choice has a price. The exam rewards matching the pricing model to the usage pattern, moving data to cheaper tiers as it cools, and avoiding unnecessary data transfer charges. Cost optimization is not about the cheapest option but the most cost-effective one that still meets the security, resilience and performance requirements set by the other domains.

How it's tested

SAA-C03 questions ask for the MOST cost-effective design that still meets the stated requirements. In this domain they describe a usage pattern and ask which pricing model, storage tier or network option minimizes cost without breaking security, resilience or performance.

  • Match the pricing model

    A usage pattern is described, and you choose On-Demand, Reserved, Savings Plans or Spot to fit it.

  • Tier the storage

    A data access pattern over time is described, and you choose lifecycle policies or Intelligent-Tiering to move data to cheaper classes.

  • Cut network cost

    A data-transfer or gateway pattern is described, and you choose VPC endpoints or placement to avoid unnecessary charges.

Reasoning it through

A workload runs steadily twenty-four hours a day for years on a fixed number of instances. Paying On-Demand is simplest but most expensive for steady use. A Savings Plan or Reserved Instances commits to that usage for one or three years at a large discount, which is the cost-effective choice for predictable, steady workloads, while Spot suits only interruptible work.

Task statements 4

Official task statements from the Amazon Web Services outline, in outline order.

4.1 Design cost-optimized storage solutionsS3 storage classes, lifecycle policies, Intelligent-Tiering and right-sized volumes.
What it means
This covers choosing S3 storage classes by access frequency, lifecycle policies that transition and expire objects, S3 Intelligent-Tiering for unknown patterns, right-sizing EBS volumes, and matching storage to cost and access needs.
What to understand
S3 classes trade retrieval cost and speed against storage cost: Standard for frequent access, Infrequent Access for cooler data, and Glacier tiers for archives. Lifecycle policies automate moving data to cheaper classes and deleting it when expired. Intelligent-Tiering moves objects automatically when the access pattern is unknown or changing.
How it shows up
A data access pattern over time is described, and you must choose the storage class, lifecycle policy or Intelligent-Tiering.
Common confusion
Using S3 Standard for rarely accessed data instead of a cheaper class or a lifecycle policy that transitions it.
4.2 Design cost-optimized compute solutionsEC2 pricing models, right-sizing, and serverless cost.
What it means
This covers EC2 pricing models, On-Demand, Reserved Instances, Savings Plans and Spot Instances, right-sizing instances, using auto scaling to match capacity to demand, and serverless cost models that charge per use. It matches the pricing model to the workload's usage pattern.
What to understand
On-Demand suits short-term or unpredictable use. Reserved Instances and Savings Plans discount steady, predictable usage in exchange for a one- or three-year commitment. Spot Instances offer deep discounts for interruptible workloads. Right-sizing and auto scaling avoid paying for idle capacity, and serverless charges only for actual use.
How it shows up
A usage pattern is described, and you must choose the pricing model and sizing approach that minimizes cost while meeting needs.
Common confusion
Using Spot Instances for workloads that cannot tolerate interruption.
4.3 Design cost-optimized database solutionsDatabase engine and capacity choices for cost efficiency.
What it means
This covers choosing cost-effective database options: the right engine and instance size, Reserved capacity for steady databases, serverless database options such as Aurora Serverless for variable load, and using read replicas or caching to avoid over-provisioning. It matches database cost to the workload.
What to understand
A right-sized database on reserved capacity is cheaper for steady load, while serverless database options scale to zero or near-zero for intermittent use. Caching and read replicas can be cheaper than scaling a large primary. Purpose-built databases can be more cost-effective than forcing one engine to do everything.
How it shows up
A database usage pattern is described, and you must choose the engine, capacity model or scaling approach that is most cost-effective.
Common confusion
Running an always-on provisioned database for an intermittent workload that a serverless option would serve more cheaply.
4.4 Design cost-optimized network architecturesData transfer, NAT gateways, VPC endpoints and placement to reduce cost.
What it means
This covers network cost factors: data transfer charges between Regions, Availability Zones and to the internet, the cost of NAT gateways, using VPC endpoints to avoid internet data-transfer charges, and placement choices that reduce cross-zone or cross-Region transfer.
What to understand
Data transfer is a major and often overlooked cost. Keeping traffic within an Availability Zone and using VPC endpoints to reach AWS services privately avoids internet and NAT charges. NAT gateways incur hourly and per-gigabyte costs, so routing matters. Content delivery and caching can also cut data-transfer cost.
How it shows up
A data-transfer or gateway cost pattern is described, and you must choose VPC endpoints, placement or routing to reduce it.
Common confusion
Overlooking data-transfer and NAT gateway charges when comparing designs.

High-yield concepts

The terms, rules and formulas this domain turns into questions most often.

  • EC2 pricing models

    On-Demand, Reserved Instances, Savings Plans and Spot.

    Remember
    Match the model to the usage pattern.
    Tested as
    Compute cost questions.
    Trap
    Using On-Demand for steady long-term workloads.

  • Reserved Instances vs. Savings Plans

    Both discount committed usage; Savings Plans are more flexible across services.

    Remember
    Commit for steady, predictable usage.
    Tested as
    Commitment pricing questions.
    Trap
    Committing for unpredictable workloads.

  • Spot Instances

    Deep discounts for interruptible capacity that AWS can reclaim.

    Remember
    Use only for fault-tolerant, interruptible work.
    Tested as
    Compute cost questions.
    Trap
    Using Spot for workloads that cannot be interrupted.

  • S3 storage classes

    Standard, Infrequent Access and Glacier tiers trading retrieval against storage cost.

    Remember
    Match the class to access frequency.
    Tested as
    Storage cost questions.
    Trap
    Keeping cold data in Standard.

  • S3 lifecycle policies

    Rules that transition objects to cheaper classes and expire them.

    Remember
    Automate moving data as it cools.
    Tested as
    Storage cost questions.
    Trap
    Managing transitions manually.

  • S3 Intelligent-Tiering

    Automatically moves objects between tiers based on access.

    Remember
    Use it when the access pattern is unknown or changing.
    Tested as
    Storage cost questions.
    Trap
    Using it where a simple lifecycle policy suffices.

  • Right-sizing

    Matching instance or resource size to actual need.

    Remember
    Avoid paying for idle capacity.
    Tested as
    Cost reduction questions.
    Trap
    Over-provisioning for peak that rarely occurs.

  • Aurora Serverless

    A database capacity model that scales with load and down when idle.

    Remember
    It suits variable or intermittent database workloads.
    Tested as
    Database cost questions.
    Trap
    Running always-on provisioned for intermittent load.

  • Data transfer cost

    Charges for moving data across zones, Regions and to the internet.

    Remember
    It is a major, often overlooked cost.
    Tested as
    Network cost questions.
    Trap
    Ignoring transfer charges when comparing designs.

  • VPC endpoints

    Private connectivity to AWS services without internet or NAT.

    Remember
    They avoid internet data-transfer and NAT costs.
    Tested as
    Network cost questions.
    Trap
    Routing AWS-service traffic through a NAT gateway unnecessarily.

  • NAT gateway cost

    NAT gateways charge hourly and per gigabyte processed.

    Remember
    Minimize traffic through them where possible.
    Tested as
    Network cost questions.
    Trap
    Sending high-volume AWS-service traffic through NAT.

Side by side

Concepts candidates mix up, compared on the point that decides the answer.

On-Demand vs. Reserved/Savings Plans vs. Spot

ConceptMeaningKey differenceCommon trap
On-Demand Pay per use with no commitment Flexible, most expensive for steady use Using it for steady long-term workloads
Reserved / Savings Plans Committed usage at a discount Cheaper for predictable steady load Committing for unpredictable workloads
Spot Spare capacity at deep discount Can be reclaimed, for interruptible work Using it for non-interruptible workloads

S3 class and lifecycle choices

ConceptMeaningKey differenceCommon trap
S3 Standard Frequent-access storage Highest storage cost, cheapest retrieval Storing cold data here
Infrequent Access / Glacier Cooler and archival tiers Cheaper storage, costlier or slower retrieval Using Glacier for frequently accessed data
Intelligent-Tiering Automatic tiering by access Best for unknown patterns Using it where a lifecycle rule suffices

VPC endpoint vs. NAT gateway for AWS-service traffic

ConceptMeaningKey differenceCommon trap
VPC endpoint Private path to AWS services Avoids internet and NAT data charges Not using it for AWS-service access
NAT gateway Outbound internet for private subnets Charges hourly and per gigabyte Routing AWS-service traffic through it

Exam traps

Each trap is an answer that looks right. Learn the clue that should change your mind.

  1. On-Demand for steady load

    What candidates pick
    Pay On-Demand for a workload that runs steadily for years.
    Why it looks right
    It needs no commitment and is simple.
    Why it's wrong
    Savings Plans or Reserved Instances discount predictable steady usage substantially.

    Clue to watch for The workload runs continuously and predictably over a long period.

  2. Spot for critical work

    What candidates pick
    Run a workload that cannot be interrupted on Spot Instances.
    Why it looks right
    Spot is the cheapest option.
    Why it's wrong
    AWS can reclaim Spot capacity, so it suits only fault-tolerant, interruptible workloads.

    Clue to watch for The workload must not be interrupted or lose state.

  3. Standard for cold data

    What candidates pick
    Keep archival data in S3 Standard.
    Why it looks right
    Standard is the default class.
    Why it's wrong
    Cooler tiers or a lifecycle policy that transitions data cut storage cost sharply for infrequently accessed data.

    Clue to watch for The data is accessed rarely or is an archive.

  4. Always-on DB for intermittent load

    What candidates pick
    Run an always-on provisioned database for an intermittent workload.
    Why it looks right
    Provisioned is the familiar model.
    Why it's wrong
    A serverless database such as Aurora Serverless scales down when idle, costing less for variable load.

    Clue to watch for The database load is intermittent or highly variable.

  5. NAT for AWS services

    What candidates pick
    Route traffic to AWS services from private subnets through a NAT gateway.
    Why it looks right
    NAT provides outbound access.
    Why it's wrong
    A VPC endpoint reaches AWS services privately and avoids NAT and internet data-transfer charges.

    Clue to watch for Private resources access AWS services and cost matters.

  6. Ignore data transfer

    What candidates pick
    Compare two designs on compute cost alone.
    Why it looks right
    Compute is the obvious cost.
    Why it's wrong
    Data transfer across zones, Regions and the internet can dominate cost and must be included.

    Clue to watch for The designs differ in cross-zone, cross-Region or internet traffic.

Answer framework

For any Domain 4 cost scenario, use these steps to find the answer SAA-C03 expects.

  1. Confirm the security, resilience and performance requirements first, then seek the most cost-effective option that still meets them.
  2. Match the compute pricing model to the usage pattern: Savings Plans or Reserved for steady load, Spot for interruptible work, On-Demand for short or unpredictable use.
  3. Tier and lifecycle storage so data moves to cheaper classes as it cools, using Intelligent-Tiering when the pattern is unknown.
  4. Right-size resources and use serverless or auto scaling to avoid paying for idle capacity.
  5. Account for network costs, using VPC endpoints and placement to cut data-transfer and NAT charges rather than comparing compute cost alone, and document the trade-off where a cheaper option would weaken a requirement.

Practice question

A company runs a production workload on a fixed number of EC2 instances continuously, with steady and predictable usage expected for the next three years. Which pricing approach is MOST cost-effective?

Want more like this? The AWS SAA practice test has questions for every domain.

Checklist

0 of 10

Tick each one you can do without notes. Your ticks stay in this browser.

Study strategy

Approach this domain as choosing the most cost-effective option that still meets the other domains' requirements, not the cheapest regardless. Start with EC2 pricing models, since they appear often: learn which usage pattern each fits, and that Spot is only for interruptible work while Savings Plans and Reserved suit steady load.

Next, learn S3 cost management: storage classes by access frequency, lifecycle policies to transition and expire data, and Intelligent-Tiering for unknown patterns. Then study the often-overlooked network costs, data transfer, NAT gateways and VPC endpoints, which decide many cost comparisons.

Use the AWS pricing calculator and a free-tier account to see how choices change cost in practice. Because cost reasoning appears across the exam, revisit the earlier domains with a cost lens. This is the final domain, so finish with full practice tests that combine security, resilience, performance and cost in single scenarios. Hands-on practice in a free-tier AWS account cements these choices far faster than reading, and reviewing missed practice questions by domain shows where to focus next.

FAQ

How much of SAA-C03 is Domain 4?

Design Cost-Optimized Architectures is 20% of the scored content, the smallest domain, though cost reasoning appears across the exam.

When should I use Reserved Instances or a Savings Plan?

For steady, predictable, long-term usage, where committing for one or three years earns a significant discount over On-Demand.

When is Spot appropriate?

Only for fault-tolerant, interruptible workloads, because AWS can reclaim Spot capacity at short notice.

How do I lower S3 storage cost?

Match the storage class to access frequency, use lifecycle policies to transition cooling data to cheaper tiers and expire it, and use Intelligent-Tiering when the access pattern is unknown.

Why do VPC endpoints save money?

They provide a private path to AWS services, avoiding internet data-transfer charges and NAT gateway costs that routing through a NAT gateway would incur.

Is cost optimization about choosing the cheapest option?

No. It means choosing the most cost-effective option that still meets the security, resilience and performance requirements of the design.

How does Domain 4 connect to the rest of the exam?

Every design choice in the other domains has a cost, so cost reasoning runs throughout, and this domain revisits storage, compute, database and network choices through a cost lens.

Other Cloud Computing certifications

Where the same subject shows up in other exams. Cards marked "Same subject" open the matching domain.

Test your Domain 4 knowledge

You have covered cost-optimized storage, compute, databases and networking. Test whether you can choose the most cost-effective option that still meets requirements under exam conditions.

Take the AWS SAA practice test Check your AWS SAA readiness

Keep going