Cloud Computing
Cloud architecture and the well-architected pillars
Cloud architecture is the craft of saying no to most of the catalogue. An architect weighs reliability, performance, security, cost and operability against each other for one workload, and the design that wins is the one whose compromises are written down and agreed. The Well-Architected pillars give those compromises names; the architect exams check whether you can make them under pressure.
Why this topic exists: Designing for reliability, performance, security and cost is what the architect exams certify (SAA-C03's four domains; AZ-305; Professional Cloud Architect) and what the Well-Architected Framework's six pillars name.
AWS organises its advice to architects into six pillars, and although the vocabulary is Amazon's, the questions apply on every platform. Microsoft and Google publish comparable frameworks of their own. The pillars are useful less as a checklist than as a reminder that improving one quality nearly always costs another.
| Pillar | The question it asks | What it usually costs |
|---|---|---|
| Operational excellence | Can the team run, observe and improve this without heroics? | Time spent on automation and runbooks before launch |
| Security | Who and what can reach the data, and how would we know? | Friction for users and developers |
| Reliability | What happens when a component, a zone or a region fails? | Duplicated resources and more complex deployments |
| Performance efficiency | Are we using the right kind and amount of resource for the load? | Ongoing measurement and tuning |
| Cost optimisation | Are we paying only for what delivers value? | Engineering effort and sometimes resilience |
| Sustainability | How much energy and hardware does this workload consume? | Occasionally speed or convenience |
How the exams frame the job
The Solutions Architect Associate exam guide states that it tests design against the Well-Architected Framework and splits its content into four domains: secure, resilient, high-performing and cost-optimised architectures. Each question describes a requirement and asks which combination of services meets it best, so the skill on trial is elimination: spotting the option that is correct but too expensive, or cheap but fragile. The expert and professional tiers change the scenario, not the principle. AZ-305 expects one design that settles identity, data, infrastructure and continuity together; SAP-C02 and Google's Professional Cloud Architect add existing systems, regulators and budgets to the stem.
Reliability has a price tag
Before designing for multiple regions, ask how long the business can be down and how much data it can lose. Many workloads are served well by several availability zones in one region, at a fraction of the effort a second region demands.
Good architects also write things down. A short decision record for each significant choice, stating the options considered, the one chosen and the pillar it sacrificed, lets a successor understand why the system looks the way it does. It also turns vague requirements into measurable ones: a recovery time, an acceptable data loss, a latency budget, a monthly spending ceiling. Without such figures, every design review becomes an argument about taste.
Where newcomers misjudge the role
The commonest mistake is treating architecture as service selection. Knowing that a queue decouples two components is exam knowledge; knowing whether this team can operate that queue at three in the morning is architecture. The second is designing for scale nobody has asked for, which buys complexity with the reliability budget. The third is skipping the review: AWS offers a free Well-Architected Tool in its console for assessing a workload against the pillars, and walking a design through such questions with the people who will run it catches more problems than any diagram. Architects usually come from development or administration for exactly this reason. They have operated the kind of system they now design, and they know which parts of it fail. The neighbouring topics on networking, storage and databases and migration supply the building blocks those decisions are made from.
Next to this topic
- Cloud fundamentalsService and deployment models, regions and availability zones, elasticity and shared responsibility: the vocabulary every foundational exam tests first (AZ-900 Cloud Concepts; CLF-C02; NIST SP 800-145).
- Cloud platforms: AWS, Azure, Google Cloud, OCIThe four platforms differ in naming, pricing, regions and certification ladders; choosing one is the first decision most learners and teams make (vendor exam guides; the four certification hubs on this site).
- Cloud governance and landing zonesAccounts, subscriptions and projects, management groups, policy, tagging, budgets and locks: how an estate is organised before anything is deployed (AZ-900 Azure management and governance, its one of the three areas, at 30 to 35 per cent; the Well-Architected security and cost pillars).
- Containers and KubernetesContainers are the unit of deployment in cloud-native work and Kubernetes the orchestrator; CNCF's Runtime and Orchestration categories and the CKA exam define the field.
- Serverless and event-driven computingFunctions, managed queues and event buses change how systems are designed and billed; SAA-C03 and the AWS Serverless knowledge badge treat it as its own discipline.
- Cloud storage and managed databasesObject, block and file storage and the managed relational and NoSQL services are the questions that reach into all four SAA-C03 domains in the associate exams (SAA-C03; AZ-104; DP-900 for the data side).
- Cloud networkingVirtual networks, subnets, load balancers, DNS, content delivery and hybrid connectivity are where on-premises networking meets the cloud (SAA-C03; AZ-104; both vendors also sell a networking specialty exam).
- Infrastructure as code and automationDeclaring environments in code (Terraform, CloudFormation, Bicep, Pulumi, Ansible) is CNCF's Provisioning category and the operating habit every cloud team is hired for (AZ-400; SAP-C02).
- Cloud operations and observabilityRunning what was built: monitoring, logging, tracing, alerting and reliability engineering (CNCF Observability and Analysis; the Well-Architected Operational Excellence pillar; the CloudOps exam, formerly SysOps).
- Cloud cost and FinOpsMeasured, variable cost is one of NIST's five cloud characteristics and the commonest failure in practice; the FinOps Framework (Inform, Optimize, Operate) and SAA-C03's cost-optimised domain make it a discipline, not an afterthought.
- Cloud migration and hybridMoving existing workloads (rehost, replatform, refactor) and running hybrid estates is a large part of enterprise cloud work, and the architect exams test it; the hosting guides cover the small end of a move.
Concepts to know
Glossary entries with the reason each one matters here.
- Load Balancing
The building block of resilience and scaling.
- Uptime & SLA
What the architecture is designed to meet, and what the provider promises.
- Caching
Performance efficiency in one word.
- CDN
Where the performance and cost pillars meet at the edge.
Certifications that test it
Vendor exams and free certificates; facts, cost and the preparation path are on each page, and the certifications hub has them all.
- AWS · SAA-C03AWS Certified Solutions Architect – AssociateThe associate architect exam.
- AWS · SAP-C02AWS Certified Solutions Architect – ProfessionalThe professional architect exam.
- Microsoft · AZ-305Microsoft Certified: Azure Solutions Architect ExpertAzure's architect exam.
- Google Cloud · PCAGoogle Cloud Professional Cloud ArchitectGoogle Cloud's architect exam.
Tools of the trade
- Amazon Web Services
The Well-Architected Framework is its house style.
Frequently asked
- Do I need to be a developer to become a cloud architect?
- Not necessarily, but you need to have built or run real systems. Most architects come from development, system administration or networking, and the associate exam itself recommends at least a year of hands-on design experience on the platform.
- Is the Well-Architected Framework only for AWS?
- The pillars are AWS's, but the questions behind them are platform-neutral, and Azure and Google Cloud publish similar frameworks. Learning one set of pillars well makes the others easy to read.
- Which architect exam should I take first?
- The associate architect exam on the platform you use. On AWS that is SAA-C03, which has no prerequisite. On Azure the architect credential requires the Azure Administrator certificate, so the route begins with AZ-104.
- How is multi-AZ different from multi-region?
- Multi-AZ spreads a workload across isolated data centres in one region and protects against a local outage with modest effort. Multi-region copies it to a distant area, which protects against a regional disaster but brings data replication, latency and failover questions that add a great deal of design work.
Courses in the directory
85 courses are filed here; the top 6 by our ranking, details and the provider link on each course page.
Last reviewed 26 September 2026 · Getting Digital
