Skip to content
Kuza Ndoto Academy

The Well-Architected Framework

Beginner → IntermediateChapter 7 · Page 27 of 303 min readUpdated 15 Nov 2026

The AWS Well-Architected Framework is AWS's collection of best practices for designing and running systems in the cloud. It organises good design into six pillars and gives you questions to review a workload against. It is also a great mental checklist for exams, interviews and your own projects.

The six pillars

AWS Well-Architected Framework Operationalexcellence run, improve Security protect Reliability recover Performanceefficiency right resources Costoptimisation spend wisely Sustainability less impact Foundation: design principles, regular reviews, the Well-Architected Tool
Figure 1. The six pillars hold up a well-architected workload. Neglecting one, such as security or cost, weakens the whole design.
PillarCore questionExample practices on AWS
Operational excellenceCan we run, monitor and improve the system smoothly?Infrastructure as code, small frequent deployments, CloudWatch alarms, runbooks, learning from incidents
SecurityAre data and systems protected?Least-privilege IAM, MFA, encryption with KMS, CloudTrail and GuardDuty, private subnets
ReliabilityDoes it keep working and recover from failure?Multiple AZs, Auto Scaling, automated backups, tested restores, health checks
Performance efficiencyAre we using the right resources, efficiently?Choosing suitable instance types, serverless, caching with CloudFront, load testing
Cost optimisationAre we delivering value at the lowest sensible cost?Budgets, tagging, right-sizing, Savings Plans, deleting idle resources
SustainabilityAre we minimising environmental impact?High utilisation, efficient instance types such as Graviton, storing less data, choosing Regions thoughtfully

General design principles

  • Stop guessing capacity: scale with demand instead of buying for the peak.
  • Test at production scale: create a full copy of the environment for testing, then delete it.
  • Automate: use IaC and pipelines so changes are consistent and repeatable.
  • Design for failure: assume any component can fail, and make sure the system survives it.
  • Make small, reversible changes: easy to roll back if something goes wrong.
  • Use data to decide: measure with metrics rather than relying on opinions.

Reviewing a workload: a worked example

Example

A team runs a county bursary application portal on one EC2 instance with a database installed on the same server, in one AZ, deployed by copying files over SSH. Applications open for two weeks a year.

PillarFindingImprovement
Operational excellenceManual deployments, no alarmsDeploy with CloudFormation or CDK through a pipeline; add CloudWatch alarms on errors and latency
SecuritySSH open to 0.0.0.0/0; database on a public serverUse Session Manager; move the database to RDS in private subnets
ReliabilityOne server in one AZ; untested backupsAuto Scaling group across two AZs behind an ALB; RDS Multi-AZ with automated backups; test a restore
Performance efficiencySlow during the application rushScheduled scaling before applications open; CloudFront for static files
Cost optimisationLarge server running all year for a two-week peakScale down out of season; consider serverless; tag and budget
SustainabilityIdle capacity most of the yearScale to demand; Graviton instances; lifecycle rules for old uploads

Notice the trade-offs. Multi-AZ improves reliability but costs more; the team must decide how much downtime is acceptable for a portal that matters most for two weeks. Well-Architected does not give one right answer; it makes the trade-offs visible.

Tools and lenses

  • The AWS Well-Architected Tool in the console walks you through the review questions for a workload, records your answers and produces an improvement plan.
  • Lenses add questions for specific kinds of workload, such as the Serverless Lens, the SaaS Lens and lenses for machine learning and generative AI.
  • The framework's whitepapers are free on the AWS website and are excellent reading for certification exams.

Key points

  • The six pillars are operational excellence, security, reliability, performance efficiency, cost optimisation and sustainability.
  • Design principles include stop guessing capacity, automate, design for failure and make small reversible changes.
  • A review compares a workload against each pillar and produces prioritised improvements.
  • Pillars involve trade-offs; the framework helps you make them deliberately.
Up next · New chapter: Practice & CareersProject: Serverless APIBuild an API with Lambda, API Gateway and DynamoDB.
→

Tip: use the ← and → keys to move between pages.