Moving from Azure administration to solution architecture is not simply learning more services. An administrator inherits a required environment and ensures that its resources behave correctly. An architect helps decide what that environment should be, which compromises are acceptable and how stakeholders will know the solution meets its obligations. Operating experience makes those decisions more realistic, but a successful deployment is not by itself evidence of a sound design.
The relationship between Microsoft AZ-104 Exam and Microsoft AZ-305 Exam reflects that shift. AZ-104 maps to Microsoft Certified: Azure Administrator Associate. AZ-305 measures designing Azure infrastructure solutions; Microsoft also requires the Administrator Associate credential for Azure Solutions Architect Expert. These formal requirements are distinct from career readiness: passing an exam does not automatically establish that someone has led a resilient, secure and cost-aware design.
Replace “which setting?” with “which requirement?”
Imagine that a mid-sized publisher operates a subscription service. It has an Azure web application, user identities, a relational subscription database, document storage and data feeds from partners. Administrators keep these components running. The business now wants expansion into another region, stronger privacy boundaries and a way to recover from an extended outage without surprising customers with duplicate charges.
An administrator may know how to configure a storage firewall, scale a compute plan, assign a role and restore a database backup. The architect must first learn which data can cross regions, how much downtime is tolerable, who owns each dependency and how to validate subscription consistency after failover. Choosing services before those constraints are clear creates expensive rework.
Translate the business request into measurable statements. “Highly available” could mean that checkout must continue during one zone failure while editorial reporting may tolerate a longer interruption. “Secure” could require specific customer-data separation and key-management responsibilities. “Cost efficient” may mean predictable baseline spending plus justified burst capacity. Those conditions are the basis for comparing Azure options rather than a list of products to deploy.
Scope shifts from an administrative boundary to a governance design
AZ-104 work includes using resource groups, subscriptions, Microsoft Entra identities, role assignments and Azure Policy to implement management boundaries. Administrators must distinguish identity authentication from Azure resource authorization and recognize how assignments inherit or are constrained by scope.
At the architecture level, management groups, subscriptions, resource groups, policies and role design express organizational accountability. A subscription boundary may serve billing, regulatory or delegated-management goals. Placing every resource in one subscription can simplify a prototype while complicating incident isolation and cost ownership for a multi-region production service.
The architect should describe who deploys resources, who reviews policy exceptions and who has permission to change customer-facing security rules. Azure Resource Manager establishes a common deployment and control-plane model, but it cannot decide the company’s division of duties. A technically valid template can still be rejected by a necessary organization policy.
Design storage from recovery and data-access obligations
An administrator can create storage accounts, configure redundancy and lifecycle rules, limit data-plane access and diagnose permissions or connectivity. Those are necessary skills. An architect must decide which data model and access pattern fits each business use: transactions may need relational consistency, product images may need object storage, and archival records may have retention constraints.
Consider invoice generation. Copying invoices to another region may improve availability, but the organization must specify which copy is authoritative and how to avoid mismatched invoices after a failover. Replication does not guarantee an application-consistent recovery point. A backup can restore data but may be too slow for the required service, and a secondary data store may be current but replicate a bad deletion.
Use Azure storage service choices to frame durability, latency, access protocols and management responsibilities. The architecture recommendation should also include application identity, private networking, encryption, retention and how a team will prove that restricted documents remain restricted after a migration.
Ask what happens when an authorized application is compromised. Storage encryption alone cannot stop that application from reading data it has permission to access. The design needs least-privilege access, logging and response controls appropriate to the threat and the underlying business record.
Compute selection becomes a lifecycle trade-off
At the administration level, configuring virtual machines, App Service, containers and relevant deployment settings requires understanding their operational behavior. A correctly running VM is an administrative success; it may still be an unnecessarily expensive or fragile hosting choice for a stateless web workload.
The architect compares management overhead, scaling behavior, portability, runtime requirements, startup constraints and failure domains. A managed service may reduce patching obligations while imposing limitations on networking or runtime customization. A container platform may improve deployment consistency but introduce orchestration complexity the team cannot currently operate.
For the publisher, separate payment state from the web interface. A stateless front end can scale differently from a stateful database. Queueing selected asynchronous work may keep customer interactions responsive during partner-system slowdowns. The architecture must specify which operations can be retried without duplicate side effects, especially where money or account changes are involved.
Networking changes from reachability to service boundaries
An administrator should be able to establish the route from client to service, check DNS, understand NSG rules and diagnose a private endpoint. An architect decides how public and private traffic should be separated, how regional failures affect application delivery and how third-party or on-premises connectivity fits the security and continuity model.
The network design must follow actual application dependencies. Restricting a database to private access can be appropriate, but a poorly planned DNS or failover path may make the database unreachable from the emergency operating region. The solution needs tests of permitted and denied connections from the relevant workloads, not just a diagram with a private-network icon.
The existing comparison of Azure networking responsibilities shows where administration ends and specialist network engineering begins. An architect must understand specialist recommendations well enough to evaluate their consequences, without pretending to replace the engineer responsible for their implementation.
Monitoring should demonstrate business outcomes
An administrator may configure Azure Monitor, diagnostic settings, alerts, log retention and recovery operations. The architect decides what “healthy” means for the complete subscription service: can a customer sign in, complete checkout, receive the correct charge and view a purchased document within the required time?
An application can return HTTP success while an asynchronous billing event never reaches its destination. A database metric may look healthy while authorization is leaking another subscriber’s records. Observability therefore needs service-level indicators, correlation across dependencies and safeguards for sensitive telemetry. The Azure monitoring concepts are useful when tied to a concrete question that operators can answer.
Operational response must be designed too. Which alerts wake the team? Which degradation triggers a failover? Who can authorize a risky rollback? Where is the recovery runbook stored if the primary region is inaccessible? A promise of monitoring without a plausible response process is an incomplete architecture.
Recovery turns resource features into business commitments
Recovery time and recovery point objectives describe different promises. The first concerns acceptable restoration delay; the second concerns tolerable loss of data changes. Availability zones, geo-redundancy, backup and disaster recovery each cover different failures and should not be treated as synonyms.
For the publisher, a design could preserve access to already purchased content during a failure while temporarily queuing new purchases. Another could prioritize continuous checkout and accept additional cost and complexity. Neither is automatically best. The architect should compare consequences for customers, data reconciliation, compliance and support staffing.
Test a scenario where a regional outage occurs during a payment write. Which system owns the transaction? How is a retry identified? Can the application confirm whether the charge occurred before accepting another one? These questions turn Azure continuity mechanisms into business correctness rather than a slide about regional resilience.
Cost governance should be included in the same evidence trail. Suppose the publisher’s second-region deployment uses a premium networking service with a fixed ongoing cost. An architect should show which recovery or performance requirement justifies that expense, what less costly alternative was considered and how the team will detect when the assumed usage changes. A deployment that meets technical acceptance while consuming substantially more than the approved budget is not automatically a satisfactory design.
Data movement also has an operational and governance cost. Moving a subscriber database between regions may trigger retention, access and contractual questions beyond the network charge. The architecture proposal should identify which party owns those decisions, which migration checks preserve consistency and what conditions would stop or reverse the change. That level of evidence separates a design recommendation from a configuration preference.
Show architecture readiness with decisions and evidence
A candidate moving beyond AZ-104 should write architecture decision records for two alternatives, not only deploy the chosen service. Each record states the requirement, constraints, options, rationale, risks, operating owner and a test of success. Test the assumptions through small controlled failures: lost dependency, denied permission, broken private DNS, stale data or a forced recovery.
Microsoft’s AZ-305 guide covers design identity, governance and monitoring; storage; business continuity; and infrastructure. Use these categories to find gaps in your decision-making while continuing to respect the practical administration skills that support them. A person can prepare for architecture work without claiming to have completed an expert credential before the formal certification requirements are met.
The strongest transition from AZ-104 to AZ-305 is the ability to explain why a resource configuration is part of a workable system. Administration proves that the individual controls can be operated. Architecture argues that those controls, taken together, meet a business obligation—and specifies how to detect when that argument stops being true.