In AZ-305 migration questions, the right strategy is the least disruptive one that still meets the stated requirements. Rehost is the fast like-for-like move when the workload is stable and does not need near-term modernization. Refactor is the code-level improvement path when technical debt, SDK changes, or cloud optimization justify development work. Rearchitect is the bigger design change when the current architecture blocks scale, agility, service boundaries, or cloud-native operations.
Microsoft updated the English version of the AZ-305 exam on April 17, 2026, and the AZ-305 exam page lists Design infrastructure solutions at 30-35%, the highest of the four weight bands on that page. The AZ-305 study guide, with skills measured as of April 17, 2026, measures this decision directly. It lists exactly this kind of work: evaluating a migration solution that leverages the Microsoft Cloud Adoption Framework for Azure, evaluating on-premises servers, data, and applications for migration, and recommending a solution for migrating workloads to infrastructure as a service (IaaS) and platform as a service (PaaS).
The eight strategies and the constraint that selects each
The current Cloud Adoption Framework strategy page (last updated July 6, 2026) defines eight strategies, each with a business driver and the key indicators that signal when to apply it. Learn the eight strategies, and match each scenario's constraints against the indicators of the strategy that fits.
| Strategy | What it changes | When the question points to it |
|---|---|---|
| Retire | Decommission the workload; nothing moves | The workload is redundant, unused, or a duplicate |
| Retain | Nothing changes for this planning cycle | The question gives a reason to stay: recent hardware, an open compliance constraint, hardware or latency lock-in |
| Rehost | Move the workload as-is to Azure VMs (lift and shift) | A stable app with compatibility constraints (vendor DLL, OS agent, custom configuration) and a deadline |
| Replatform | Move to a managed platform with minimal code change | The operations burden is the stated problem and a clean managed equivalent exists |
| Refactor | Change code or configuration to fit the cloud, inside the existing design | The question names technical debt, SDK updates, performance, or observability work |
| Rearchitect | Redesign the architecture with cloud-native patterns (microservices, serverless, event-driven) | The current design blocks scale, agility, or service boundaries |
| Rebuild | Redevelop the application from scratch | The app is beyond repair and the question requires a full rewrite |
| Replace | Swap the application for a SaaS product | The question names an off-the-shelf SaaS equivalent that covers the need |
| Verdict | Match the stated constraint to the strategy that fits it | The stated constraint selects how much modernization the workload needs |
The migration decision path
A later arrow in the diagram is a later phase, not a required order. The constraint in the question decides the answer, and it can point at rehost or at rearchitect. A correct answer is rehost when the scenario rewards speed, risk control, and compatibility. A correct answer is rearchitect when the existing design cannot meet the target scale or operating model.
What the Cloud Adoption Framework adds
The strategy page lays out a process for choosing among the eight strategies. It starts from business drivers: a business driver defines why a workload needs to change, and each strategy pairs that driver with the key indicators that signal when to apply it. For AZ-305, use that process in three ways:
- Ask whether the workload should move at all before choosing a target service.
- Separate migration path from hosting target. A VM target does not automatically mean the best strategy was chosen.
- Match modernization depth to the stated constraint. If the scenario only asks for datacenter exit speed, rehosting is enough, and a rewrite adds cost the requirement never asked for.
Worked scenario: the payroll portal deadline
A company has a payroll portal running on three Windows Server VMs in VMware. The app uses IIS, a scheduled Windows service, a SQL Server database, and a vendor DLL. The datacenter lease expires in six months. The business requirement is to keep payroll stable through year-end, reduce datacenter dependency, and avoid large code changes before the next compliance audit. The architecture team also wants a future path to managed web hosting and better release automation.
The answer is rehost now. The deadline, the vendor DLL, and the audit risk make a large redesign hard to justify, and the constraints in the question reward stability over a redesign. Rehost the app servers to Azure VMs if the Azure Migrate assessment confirms readiness, sizing, cost, and dependencies. Azure Migrate discovers workloads, assesses whether on-premises servers, SQL servers, and web apps are ready for migration, estimates right-sizing and cost, and maps the network dependencies between servers (Azure Migrate overview).
The managed-hosting goal becomes a later replatform or refactor phase. Azure Migrate also assesses on-premises web apps for App Service. If the vendor DLL, the scheduled Windows service, or a registry or COM dependency blocks normal App Service hosting, the assessment data says the workload stays on VMs. The managed-hosting goal stays on the roadmap for components that fit it.
Rearchitect only if the scenario changes. If payroll must scale calculation, approval, notification, and reporting independently, or the monolith blocks release speed and resilience targets, a phased rearchitect path becomes defensible. Microsoft's modernization guidance defines rearchitecting as redesigning application architecture using cloud-native patterns, microservices, serverless, and event-driven (Plan cloud modernization). Rearchitect is on the table when the question's problem is the design itself; a deadline alone points at rehost.
Exam-ready answer: rehost now, plan a later replatform or refactor for the web tier if the assessment proves it is compatible, and reserve rearchitecting for a later phase with explicit scalability or service-boundary requirements.
Service boundaries that change the answer
Azure Migrate helps you decide on, plan, and execute the migration. The strategy still comes from the workload's constraints, because that is what the exam scenario tests.
Use Azure VMs when OS control, installed agents, legacy software, custom configuration, or lift-and-shift compatibility are real constraints. The compute decision tree describes IaaS as the most similar to a traditional on-premises environment. Microsoft manages the infrastructure, and you manage the VMs (Choose an Azure compute service).
Use App Service when the workload is a web app or API that fits a managed web platform. Use Container Apps when containerized workloads matter without direct Kubernetes API control. The comparison page says AKS supports direct access to the Kubernetes API, while Container Apps gives a fully managed experience with scale to zero, revisions, and traffic splitting. Use AKS when Kubernetes API or control-plane access is an explicit requirement (Container Apps comparison). Use Functions when the unit of work is event-driven code that runs to completion per event.
If the scenario is about preserving VMware operations and platform compatibility, Azure VMware Solution becomes part of the design. Microsoft describes it as private clouds that contain VMware vSphere clusters built from dedicated bare-metal Azure infrastructure, and it is the right fit when the question asks to keep vSphere tooling and operations with minimal retraining (Azure VMware Solution overview).
For adjacent compute placement decisions, the AZ-305 compute hosting guide is the better next read. If the question shifts to private connectivity, datacenter reachability, or DNS, use the AZ-305 hybrid network connectivity guide.
What wrong answers reveal
- Rearchitect for the payroll deadline spends time on a redesign the constraint does not ask for. The deadline, the vendor DLL, and the audit risk all reward stability over a redesign.
- Rehost for a fragile, unscalable monolith moves the workload but not the fragility. The failure mode the business needs fixed stays with the app.
- App Service because the workload has IIS skips the dependency check. Installed agents and unsupported configuration do not run on a managed web platform.
- AKS because containers are mentioned misses the simpler fit. Container Apps covers the serverless container pattern without Kubernetes API control.
- Azure VMware Solution for every VMware source assumes the requirement is to keep vSphere. The platform fits when the scenario asks to keep vSphere operations and tooling.
- Migrating every workload leaves out the first decision, which is whether to move at all. Retain, retire, replace, and rebuild are valid business decisions in their own right.
A practice test or diagnostic shows which migration decisions are weak. When the weak areas are timing, compatibility, operating model, code debt, architecture, or target-service fit, work those areas next.
How to study this for AZ-305
Practice by changing one requirement at a time:
- Replace "six-month datacenter deadline" with "must support 3x traffic and independent release cycles." Does rearchitect become stronger?
- Add "vendor agent must be installed on the OS." Does rehost to VMs become defensible?
- Remove the vendor dependency and add "reduce patching and hosting operations." Does replatform to App Service become stronger?
- Add "queue-triggered document processing." Do Functions or Container Apps jobs become part of the design?
- Add "keep VMware tooling and vSphere operations with minimal retraining." Does Azure VMware Solution become worth evaluating?
When you want to see whether this decision pattern is costing you points, take the AZ-305 assessment. It costs nothing and needs no credit card, and it hands back a Readiness Report that lists your top gaps. That report is your starting point. As you keep practicing on CramHQ, CramHQ adds the gaps your work exposes to the ones in the Readiness Report, weighs how many points each gap would take off your score, and leads the plan with the repair worth the most points. The plan is the Pass Plan: your ranked list of targeted repair tasks and the core of the product. For AZ-305, the path is US$34.99 for 12 months of access, and the first repair is free. The practice tests and the timed simulation check your readiness while you work through the plan. Start with the gaps in the Readiness Report to find which migration decision is weak, work the repairs in order, and retest the nearby scenarios that test the same decision.