Topic lessonTopic: AWS design through six pillars
AWS Well-Architected explained: design trade-offs across six pillars
AWS Well-Architected is a framework for reviewing how you design and operate a system through six pillars. Adding an application instance, restricting access or reusing a result can bring benefits and new constraints. Follow one report application and separate the intended improvement, the trade-offs and the evidence still needed.
Turn the six pillars into practical design questions. Explain a change’s benefits, trade-offs and evidence needs, distinguishing a violated requirement from an outcome that remains unverified.
For readers beginning to think of AWS as a combination of services. An application processes user requests; a workload includes the application, data and operations that serve a purpose. The example is explained here. No AWS account or hands-on operation is required.
Start with one report application and a shared set of requirements
A pillar is a design perspective, not a service category or a step in an execution sequence. Use AWS Well-Architected to discuss the advantages and disadvantages of decisions and identify improvements. Checking six boxes does not certify a system as secure.
- Reader → Report appRequest a report
- Report app → Saved reportsRetrieve the required content
- Reader
- An authorized staff member reads the required morning report.
- Report app
- Checks the reader, selects the permitted report and returns its contents.
- Saved reports
- Contains prepared results and source records that must be retained.
The lines show request and data relationships. They do not depict a real network path, an AWS deployment or the complete authentication process. Keep the application’s purpose and required data fixed when comparing later changes.
| Requirement | What is given |
|---|---|
| Who can read | Only authorized staff may read the reports they need. |
| Required freshness | The morning report must include approved changes through 07:00. |
| Data retention | Keep source records for 30 days. Recreating a prepared report is separate from retaining its source. |
| Failure preparation | Decide what must be recovered and how to check report reads if the application becomes unavailable. No recovery-time limit is specified here. |
| Comparison limits | Prices, response times and emissions have not been measured. Actual improvement from a deployment change is unverified. |
Translate the six pillars into practical questions
The six pillars are operational excellence, security, reliability, performance efficiency, cost optimization and sustainability. A change can affect several pillars; they do not describe six separate systems to build.
| Pillar | Focus | Question for this application |
|---|---|---|
| Operational excellence | Improve change, observation and response to keep delivering the required outcomes. | Can we identify the changed version and owner, detect problems and improve the procedure? |
| Security | Protect data, systems and assets. | Who may read what? Does that role also need permission to change or delete it? |
| Reliability | Provide the intended function correctly when it is needed. | How will staff regain the required report if the app or a dependency fails? |
| Performance efficiency | Match resources and approaches to performance requirements. | Can the app return the required correct result quickly enough when demand is concentrated? |
| Cost optimization | Examine spending against the value the workload must deliver. | Can we remove unnecessary usage while meeting read, freshness, retention and recovery requirements? |
| Sustainability | Examine resource use and environmental impact for the useful output. | Are we recomputing the same result unnecessarily, or adding avoidable storage and transfer? |
Operations and security: distinguish making a change from delegating permission
Operational excellence considers ownership, procedures, observations and learning from problems together. Make it possible to check the required report result, not merely that an automated job finished. Small reversible changes and continual refinement of procedures are part of this work.
Security concerns protecting the workload while allowing the right identities to perform the operations they need. Least privilege means limiting permission to the necessary scope. Access restrictions, action records, data protection and incident response are related controls, not interchangeable evidence.
| Given condition or decision | What follows in this example |
|---|---|
| The report-serving role only needs to read specified reports. | Changing and deleting reports are unnecessary permissions for that role. |
| A separate role is responsible for writing prepared reports. | Review each role’s required operations rather than indiscriminately prohibiting all writes. |
| A permission change has been proposed but not applied to a real environment. | It is a least-privilege proposal, not evidence of a safe rollout or successful real read. |
| Define post-change checks. | Record the actor and change, then verify permitted reads and operations that should be rejected. |
Choose an option to see the result and explanation.
Reliability and performance: separate failure preparation from normal speed
Reliability considers whether the workload can provide its intended function correctly when expected. Test recovery procedures and examine shared dependencies. Having more application instances does not by itself guarantee the required data or continued use.
Performance efficiency considers how well technologies and resources meet the required work as demand changes. Compare the same correctness criteria, work and response-time requirements. Choosing a larger server or newer service does not establish the resulting performance.
| Check | Known and unresolved conditions |
|---|---|
| One application instance fails | The proposal provides another instance, but routing, transition, remaining capacity and data access have not been tested. |
| Both read the same storage | Additional app instances alone do not remove failure of a shared storage or identity dependency. |
| More readers arrive together | Check whether requests are distributed and correct results meet the response requirement under equivalent conditions. There are no measured results here. |
| Cost and operations | Review additional resources, rollout procedures, configuration consistency and ongoing cost too. |
Choose an option to see the result and explanation.
Cost and sustainability: preserve useful outcomes while reducing waste
Cost optimization relates expenditure to the value the business needs. Examine unnecessary usage and spending while maintaining required functionality, freshness, retention and recovery. A cheaper design that removes required outcomes is a different proposal. Estimates and actual bills are also different evidence.
Sustainability considers environmental impact and the compute, storage, transfer and other resources used for a useful outcome. Eliminating redundant work is a candidate improvement. Do not quantify energy or emissions reductions without examining the actual conditions.
| Proposal | Requirement comparison | Conclusion |
|---|---|---|
| Delete source records after 7 days instead of 30 | This violates the stipulated 30-day retention requirement. | Do not adopt it under these requirements. This is a known conflict, separate from the unknown savings. |
| Reuse a report representing 06:00 | The report must include approved changes through 07:00, but this result is stipulated to omit later changes. | The freshness requirement is not met, even if computation is avoided. |
| Reuse a 07:15 result stipulated to contain the required changes through 07:00 | The given freshness condition is satisfied. Reuse must still restrict who can read what. | A candidate for reuse; access, recreation, performance, cost and environmental impact still need validation. |
Choose an option to see the result and explanation.
Review one proposal through all six pillars
Now consider reusing the 07:15 result for the permitted readers. The input states that it includes the required changes through 07:00, so the stipulated freshness requirement is met. That does not approve the entire change.
Including required changes through 07:00 is stipulated. The checks below remain separate.
- Operational excellence
- Which version is served, and who detects and addresses an incorrect result?
- Security
- Do the recipients read only their permitted reports?
- Reliability
- If the result is lost, can it be recreated and read from retained source records?
- Performance efficiency
- Does the required correct result meet its response criteria?
- Cost optimization
- Have recomputation, storage, transfer and operations been compared?
- Sustainability
- Does less computation avoid shifting waste to storage or transfer?
| Category | Decision record for this proposal |
|---|---|
| Established by the input | The result is stipulated to include required changes through 07:00. Its 07:15 timestamp alone does not establish this. |
| Still unverified | Recipient access, recovery, response time, cost and environmental impact. A written plan is not evidence it was carried out. |
| Next action | Assign owners and scope; check allowed/rejected reads, recreated contents, equivalent-load results, resource use and cost assumptions. |
A trade-off exchanges a benefit for another burden or constraint. Reuse might avoid repeated work while adding freshness management, recipient access checks and storage operation. Do not add pillar scores to declare a winner. Record whether required conditions are met and who will decide whether remaining risks are acceptable.
Explain the decision and the evidence still needed
- Write down required functions, permitted readers, freshness, retention and failure conditions.
- Apply all six design questions to one proposed change.
- Distinguish a requirement conflict from a condition left unverified by missing evidence.
- Assign the checks and define which results would justify adoption or revision.
Choose an option to see the result and explanation.
You can now explain the same application change through six design questions. Use related lessons later to deepen change management, access control, availability and performance. Do not change a real environment solely from the configuration labels in this example.
Check, then explain in your own words
- Express the six pillars as practical questions about the report application.
- Distinguish the 06:00 and 07:15 reuse proposals using the freshness requirement.
- Record benefits, burdens, established conditions and unresolved checks for one change.
Check your understanding: Someone reports that a report is now faster and cheaper to read. Is the six-pillar review complete?
No. Check the relevant requirements and evidence for content, permitted readers, retention, failure response, operations, environmental impact and other concerns.
Improvement in one perspective does not establish the other necessary conditions or make the remaining burdens acceptable.
Sources and conditions checked
The official references used in this lesson are collected here with their scope, conditions and corresponding sections. Independent expert review and live AWS comparison have not been performed.
- AWS Well-Architected FrameworkChecked:
Scope and conditionsA framework for examining design decisions and improvements; an architecture review is not an audit or certification.
Used in this lessonStart with one report application and a shared set of requirements Translate the six pillars into practical questions Review one proposal through all six pillars
- The pillars of the frameworkChecked:
Scope and conditionsThe six pillars and their distinct concerns; no composite teaching score is an AWS assessment.
Used in this lessonStart with one report application and a shared set of requirements Translate the six pillars into practical questions
Scope and conditionsBusiness outcomes, actionable observations, small reversible changes and improving procedures; automation alone does not establish success.
Used in this lessonTranslate the six pillars into practical questions Operations and security: distinguish making a change from delegating permission
- Security: design principlesChecked:
Scope and conditionsLeast privilege, separation of duties, traceability, layered controls and data protection; one access restriction is not a security guarantee.
Used in this lessonTranslate the six pillars into practical questions Operations and security: distinguish making a change from delegating permission
- Reliability: design principlesChecked:
Scope and conditionsRecovery testing and common failure points; additional instances alone do not establish continuity or restoration.
Used in this lessonTranslate the six pillars into practical questions Reliability and performance: separate failure preparation from normal speed
Scope and conditionsAlign technology with workload requirements and compare alternatives; service names or resource counts do not establish performance.
Used in this lessonTranslate the six pillars into practical questions Reliability and performance: separate failure preparation from normal speed
Scope and conditionsRelate business output to cost and examine usage and spending; this lesson provides no live prices or bill calculation.
Used in this lessonTranslate the six pillars into practical questions Cost and sustainability: preserve useful outcomes while reducing waste Review one proposal through all six pillars
Scope and conditionsConsider resource use and environmental impact per useful output; prices or job counts alone do not quantify emissions.
Used in this lessonTranslate the six pillars into practical questions Cost and sustainability: preserve useful outcomes while reducing waste Review one proposal through all six pillars