AWS SAP · Domain 3 of 4

AWS SAP Domain 3: Continuous Improvement for Existing Solutions

This domain covers improving workloads already in production: raising operational excellence, security, performance and reliability, and finding cost-optimization opportunities, usually by changing an existing design rather than starting over.

Exam weight
25%
Official task statements
5
Estimated questions estimate
~16
Suggested study time estimate
35–45 hrs

About 10 min read Reviewed October 2026

Overview

Continuous Improvement for Existing Solutions covers observing, optimizing and evolving workloads already in production. It has five task statements, one per Well-Architected improvement area: operational excellence, security, performance, reliability and cost. Rather than greenfield design, it gives an existing architecture with a problem, slow performance, high cost, a security gap, reliability weakness or operational pain, and asks for the change that best improves it. It spans observability and automation, tightening security, tuning performance with caching and right service selection, improving resilience, and cutting cost through right-sizing, pricing models and storage tiering.

At 25%, this domain is close in weight to the others. Its questions reward incremental, targeted improvement: identifying the weakest aspect of a running system and the most effective change, often without a full rebuild. The skill is diagnosing an existing design and prescribing the highest-impact, best-balanced improvement.

How it's tested

SAP-C02 questions in this domain describe an existing workload with a weakness and ask for the best improvement. Expect the correct option to fix the stated weakness while leaving the rest of the working system intact, so discard answers that rebuild what is not broken. Prefer the targeted, highest-impact change over a full rebuild.

  • Diagnose the weakness

    A running workload's problem is described, and you identify which pillar it affects and the best fix.

  • Optimize cost

    A cost concern is described, and you choose right-sizing, a pricing model, storage tiering or architectural change.

  • Improve without rebuilding

    A limitation is described, and you choose the targeted change rather than a full redesign.

Reasoning it through

An existing workload runs steadily on On-Demand EC2 and the business wants to cut cost without changing the architecture. Rewriting to serverless is a large change the scenario does not ask for. Because the usage is steady and predictable, applying a Savings Plan or Reserved Instances cuts cost substantially with no architectural change, which is the targeted, highest-impact improvement the exam expects.

Task statements 5

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

3.1 Determine a strategy to improve overall operational excellenceObservability, automation and operational practices.
What it means
This covers improving operations of a running workload: observability with CloudWatch, X-Ray and dashboards, automation with Systems Manager and EventBridge, infrastructure as code, and operational practices that reduce manual effort and speed recovery.
What to understand
Operational excellence improvements add visibility and automation so problems are detected and remediated faster with less manual work. Automating runbooks, improving logging and tracing, and codifying infrastructure are common high-impact changes.
How it shows up
An operational pain point is described, and you must choose the observability or automation improvement.
Common confusion
Treating operational excellence as only monitoring, rather than including automation and infrastructure as code.
3.2 Determine a strategy to improve securityTightening identity, encryption and threat protection on a running system.
What it means
This covers improving the security of an existing workload: tightening IAM to least privilege, adding encryption, improving secret management, adding threat detection with GuardDuty, and closing compliance gaps without redesigning the whole system.
What to understand
Security improvements target the weakest control: over-broad permissions, missing encryption, exposed secrets or lack of detection. The best change closes the gap with minimal disruption, such as adding least-privilege roles or enabling a managed detection service.
How it shows up
A security weakness is described, and you must choose the improvement that closes it.
Common confusion
Proposing a redesign where tightening a specific control would close the gap.
3.3 Determine a strategy to improve performanceCaching, right service selection and bottleneck removal.
What it means
This covers improving performance of a running workload: adding caching with ElastiCache, DAX or CloudFront, right-sizing or changing instance and storage types, offloading reads with replicas, and removing bottlenecks identified through monitoring.
What to understand
Performance improvements target the measured bottleneck: caching for repeated reads, the right storage type for IOPS, or offloading work. The best change addresses the actual constraint rather than broadly scaling everything.
How it shows up
A performance problem is described, and you must choose the targeted improvement.
Common confusion
Scaling everything up rather than addressing the specific bottleneck with caching or the right service.
3.4 Determine a strategy to improve reliabilityAdding resilience and removing single points of failure.
What it means
This covers improving reliability of an existing workload: adding multi-AZ or multi-Region resilience, Auto Scaling, decoupling, better backup and failover, and removing single points of failure to meet higher availability targets.
What to understand
Reliability improvements find and remove the single point of failure or add the missing resilience, such as moving to multi-AZ, adding a standby, or decoupling a tight dependency, matched to the new availability target.
How it shows up
A reliability weakness is described, and you must choose the resilience improvement.
Common confusion
Adding resilience beyond the requirement, increasing cost without benefit.
3.5 Identify opportunities for cost optimizationRight-sizing, pricing models, storage tiering and architectural savings.
What it means
This covers cutting cost on a running workload: right-sizing with Compute Optimizer, choosing pricing models such as Savings Plans, Reserved Instances and Spot, S3 storage class and lifecycle optimization, reducing data transfer, and architectural changes such as serverless where they lower cost.
What to understand
Cost optimization matches spend to need: right-sizing removes waste, commitment pricing discounts steady load, Spot suits interruptible work, and storage tiering and lifecycle policies lower storage cost. The best change cuts cost while preserving the required performance and reliability.
How it shows up
A cost concern is described, and you must choose the optimization that cuts cost without breaking requirements.
Common confusion
Choosing the cheapest option regardless of the performance or reliability it sacrifices.

High-yield concepts

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

  • Observability improvement

    Adding metrics, logs, tracing and dashboards.

    Remember
    It speeds detection and diagnosis.
    Tested as
    Operational questions.
    Trap
    Treating it as monitoring only.

  • Automation

    Systems Manager and EventBridge for automated remediation.

    Remember
    It reduces manual toil.
    Tested as
    Operational questions.
    Trap
    Keeping operations manual.

  • Least-privilege tightening

    Reducing over-broad IAM permissions.

    Remember
    A common security improvement.
    Tested as
    Security questions.
    Trap
    Leaving broad permissions in place.

  • Managed threat detection

    GuardDuty and similar services.

    Remember
    Adds detection with little effort.
    Tested as
    Security questions.
    Trap
    Overlooking managed detection.

  • Caching for performance

    ElastiCache, DAX and CloudFront.

    Remember
    Targets repeated reads and content.
    Tested as
    Performance questions.
    Trap
    Scaling instead of caching.

  • Right-sizing

    Matching resources to actual usage.

    Remember
    Removes over-provisioning waste.
    Tested as
    Cost questions.
    Trap
    Leaving idle capacity.

  • Pricing models

    Savings Plans, Reserved Instances, Spot.

    Remember
    Match the model to usage.
    Tested as
    Cost questions.
    Trap
    Using On-Demand for steady load.

  • Storage tiering and lifecycle

    S3 classes and lifecycle policies.

    Remember
    Moves data to cheaper tiers as it cools.
    Tested as
    Cost questions.
    Trap
    Keeping cold data in Standard.

  • Single point of failure removal

    Multi-AZ, standby or decoupling.

    Remember
    Raises reliability.
    Tested as
    Reliability questions.
    Trap
    Adding resilience beyond need.

  • Targeted improvement

    Changing the weakest aspect, not rebuilding.

    Remember
    Highest impact for least disruption.
    Tested as
    All improvement questions.
    Trap
    Proposing a full rewrite.

Side by side

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

Targeted improvement vs. full rebuild

ConceptMeaningKey differenceCommon trap
Targeted improvement Change the weakest aspect High impact, low disruption Rebuilding unnecessarily
Full rebuild Redesign the whole workload Large effort and risk Using it where a change would do

Cost optimization levers

ConceptMeaningKey differenceCommon trap
Right-sizing Match resources to usage Removes waste Leaving idle capacity
Pricing models Commit or use Spot Discounts by usage pattern On-Demand for steady load
Storage tiering Cheaper classes as data cools Lowers storage cost Cold data in Standard

Scale vs. cache vs. optimize

ConceptMeaningKey differenceCommon trap
Scale Add capacity Handles more load Scaling the wrong bottleneck
Cache Serve repeated reads Cuts latency and cost Scaling instead
Optimize the design Change the weak component Targets the real constraint Broad changes that miss the cause

Exam traps

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

  1. Rewrite to serverless

    What candidates pick
    Rewrite a steady workload to serverless to cut cost.
    Why it looks right
    Serverless can be cheaper for some workloads.
    Why it's wrong
    For steady predictable load with no redesign requested, a Savings Plan or Reserved Instances cuts cost with no architectural change.

    Clue to watch for The usage is steady and no architectural change is requested.

  2. On-Demand for steady load

    What candidates pick
    Leave a steady production workload on On-Demand pricing.
    Why it looks right
    It needs no commitment.
    Why it's wrong
    Savings Plans or Reserved Instances discount steady usage substantially.

    Clue to watch for The workload runs continuously and predictably.

  3. Scale the wrong thing

    What candidates pick
    Scale up all tiers to fix a single slow component.
    Why it looks right
    More capacity seems to help.
    Why it's wrong
    The fix should target the measured bottleneck, often with caching or the right service.

    Clue to watch for Monitoring points to one specific slow component.

  4. Over-resilient upgrade

    What candidates pick
    Add active-active multi-Region to a workload with a modest availability target.
    Why it looks right
    More resilience seems better.
    Why it's wrong
    Resilience should match the target; over-building adds cost without benefit.

    Clue to watch for The availability requirement is modest.

  5. Cheapest regardless

    What candidates pick
    Choose the cheapest option even though it misses the performance target.
    Why it looks right
    It cuts cost most.
    Why it's wrong
    Cost optimization must preserve the required performance and reliability.

    Clue to watch for A performance or reliability requirement accompanies the cost goal.

Answer framework

For any Domain 3 scenario, use these steps to find the answer SAP-C02 expects.

  1. Diagnose which Well-Architected pillar the existing workload is weakest in, using the problem described.
  2. Prefer a targeted, highest-impact change over a full rebuild unless a rebuild is clearly required.
  3. For cost, match the lever to the cause: right-sizing for waste, pricing models for steady load, storage tiering for cold data.
  4. For performance and reliability, address the specific bottleneck or single point of failure rather than broadly scaling.
  5. Ensure the improvement preserves the other constraints, improving one pillar without breaking another. Confirm the chosen approach beats the other workable options on the stated constraints before committing to it, confirming it beats the other workable options on the stated constraints.

Practice question

An existing production workload runs on a fixed fleet of On-Demand EC2 instances with steady, predictable usage. The business wants to reduce cost without changing the application architecture. What is the MOST appropriate recommendation?

Want more like this? The AWS SAP 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 diagnosis followed by targeted prescription, not greenfield design. Practice reading an existing architecture, identifying the weakest Well-Architected pillar, and choosing the highest-impact change that does not break the others. Resist the urge to rebuild; the exam usually rewards the smallest effective change.

Give cost optimization focused attention, since it is a whole task statement and a common weak area: learn right-sizing, the pricing models matched to usage, and storage tiering and lifecycle. For performance and reliability, practice targeting the specific bottleneck or single point of failure rather than scaling broadly.

Because the scenarios resemble real operational decisions, experience and practice questions both help. Afterwards, move to Domain 4, accelerating workload migration and modernization, the final domain. For each missed improvement question, note which pillar the correct answer improved and which existing constraint your choice would have broken. Then redo the question from the starting architecture, listing the smallest change that meets the goal, because the exam favors targeted improvements over migrations or rebuilds.

FAQ

How much of SAP-C02 is Domain 3?

Continuous Improvement for Existing Solutions is 25% of the scored content, covering optimization of workloads already in production.

Should I usually recommend a full rebuild?

Usually not. The exam rewards the targeted, highest-impact change that improves the weakest aspect with the least disruption, unless a rebuild is clearly required.

How do I cut cost on a steady workload without changing the architecture?

Apply a Savings Plan or Reserved Instances to discount the steady usage, which requires no architectural change, rather than rewriting to serverless or using Spot.

How do I improve performance of an existing system?

Target the measured bottleneck, for example with caching for repeated reads or the right storage type for IOPS, rather than scaling every tier.

What are the main cost-optimization levers?

Right-sizing to remove waste, pricing models matched to usage, storage class and lifecycle optimization, and reducing data transfer, plus architectural change where it clearly lowers cost.

How is this domain different from Domain 2?

Domain 2 designs new solutions from scratch, while Domain 3 improves existing workloads, which is more about diagnosis and targeted change than greenfield design.

How does Domain 3 connect to the rest of the exam?

It improves the solutions designed in Domain 2 within the enterprise structure of Domain 1, and migrated workloads from Domain 4 often need this continuous improvement.

Other Cloud Computing certifications

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

Test your Domain 3 knowledge

You have covered improving operations, security, performance, reliability and cost of running workloads. Test whether you can choose the highest-impact targeted improvement under constraints.

Take the AWS SAP practice test Check your AWS SAP readiness

Keep going