Migrations
CAF strategies · Azure Migrate · DMS · Data Box
Practice questions for this topic →
Decision table
| Need | Choose |
|---|---|
| Discover and assess VMware, Hyper-V and physical servers, with cost estimates | Azure Migrate |
| Find which servers communicate so they move in the same wave, with no agents | Agentless dependency analysis (Azure Migrate) |
| Move quickly with no code changes | Rehost |
| Move to PaaS or containers with minimal code changes | Replatform |
| Replace a custom application with a SaaS product | Replace |
| Legacy system too outdated or inflexible; build a new cloud-native app | Rebuild |
| Stable, compliant workload with no driver to move | Retain |
| Limited business value; migration costs outweigh the benefit | Retire |
| SQL Server to SQL Managed Instance with minimal downtime | DMS online migration |
| SQL Server to Azure SQL Database with near-zero downtime | Not possible through DMS: Azure SQL Database migrations are offline only |
| Around 50 TB with limited bandwidth | Data Box |
| A few TB with limited bandwidth | Data Box Disk |
| Use your own disk drives to import to Blob Storage or Azure Files | Azure Import/Export service |
| Seed a large dataset, then keep syncing changes | Offline seed (Data Box) + incremental online transfer |
The strategy names are the Cloud Adoption Framework's. Pick the strategy from how much the app changes; pick the tool from the source platform and the bandwidth you have.
Migration strategies
| Strategy | What changes | Pick when… |
|---|---|---|
| Rehost | Nothing in the app; move the VM as-is | Speed matters and no code changes are allowed |
| Replatform | Moves to PaaS or containers with minimal code changes | You want to reduce OS and licensing overhead, e.g. IIS app to App Service |
| Refactor | Code changes to reduce technical debt, not the architecture | The app works but is hard to maintain |
| Rearchitect | The architecture: a monolith split into independently scalable services | You need cloud-native capabilities or mixed technology stacks |
| Rebuild | A new application on modern frameworks and tools | The legacy system is too outdated or inflexible to move |
| Replace | Swap the custom app for a SaaS product | A SaaS product covers the business need |
| Retain | Nothing for now | Stable and compliant, with no business driver to move |
| Retire | Switch the workload off | Little value; migrating would cost more than it delivers |
Azure Migrate and dependency analysis
- Azure Migrate discovers and assesses VMware, Hyper-V and physical servers, with cost estimates.
- Agentless dependency analysis is generally available for VMware, Hyper-V, physical and other-cloud servers, and it's Microsoft's recommended approach. The Azure Migrate appliance polls TCP connection data every five minutes.
- Agent-based dependency analysis exists only in the classic view. That view is scheduled for deprecation by the end of 2026, and you can't onboard new servers to it. Choose agentless for new discovery.
Database migration (DMS)
- SQL Server to SQL Managed Instance: online migration with minimal downtime. Managed Instance is the PaaS target for online DMS migrations.
- SQL Server to SQL Server on Azure VMs: online migration is also supported. It's IaaS, not PaaS.
- SQL Server to Azure SQL Database: offline only through DMS.
- MongoDB to Azure Cosmos DB; MySQL to MySQL Flexible Server; PostgreSQL to PostgreSQL Flexible Server.
- Assessment: SQL Server enabled by Azure Arc migration readiness assessment feeds DMS assessment reports.
- Retired: Azure Data Studio, which hosted the Azure SQL Migration extension, retired on February 28, 2026. Use the dedicated migration tooling, or VS Code with the MSSQL extension. DMS (classic) for SQL Server scenarios retired on March 15, 2026.
Offline transfer: Data Box and Import/Export
- Data Box is an offline appliance for tens to hundreds of terabytes when network bandwidth isn't enough.
- Data Box Disk ships 1 to 5 encrypted 8 TB SSDs, about 40 TB at most. Workflow: order, receive and unlock, copy the data, return the disks, and the data uploads automatically in the Azure datacenter.
- Data Box Heavy is retired.
- Seed then sync: use Data Box for the initial bulk load, then send changes over the network until cutover.
- Azure Import/Export lets you supply your own disk drives to import data into Blob Storage and Azure Files, and to export from Blob Storage.
- Keep on-prem file servers but tier cold files to Azure: Azure File Sync with cloud tiering.
Exam Tips
Tick each tip once you can answer it without looking.
“Discover and assess VMware, Hyper-V and physical servers, with cost estimates” → Azure Migrate.
“Which servers must move together, no agents” → Agentless dependency analysis. Trap: Agent-based analysis is classic-view only and being deprecated.
“Move quickly with no code changes” → Rehost.
“Move to PaaS with minimal code changes” → Replatform.
“Split a monolith into independently scalable services” → Rearchitect.
“Too outdated to move; new cloud-native app” → Rebuild.
“Replace a custom app with SaaS” → Replace.
“Stable, compliant, no driver to move” → Retain. “Little value, costly to migrate” → Retire.
“SQL Server to Managed Instance with minimal downtime” → DMS online migration.
“SQL Server to Azure SQL Database with near-zero downtime” → Not through DMS. Trap: DMS to Azure SQL Database is offline only.
“Around 50 TB, limited bandwidth” → Data Box. Trap: Data Box Disk tops out near 40 TB.
“Initial bulk load, then keep syncing” → Offline seed with Data Box, then incremental online transfer.
“Your own disk drives into Blob or Azure Files” → Azure Import/Export service.
“MongoDB, MySQL or PostgreSQL migrations” → Cosmos DB, MySQL Flexible Server, PostgreSQL Flexible Server through DMS.
“Azure Data Studio as a migration tool” → Retired February 28, 2026. Trap: Use dedicated migration tooling or VS Code with the MSSQL extension.
Sources
- Azure Migrate dependency visualization
- Azure Database Migration Service
- SQL Server to Managed Instance guide
- DMS supported scenarios
- Azure Data Box overview
- Azure Data Box Disk overview
- Azure Data Box Heavy overview
- Cloud Adoption Framework migration strategies
- What's happening to Azure Data Studio
- DMS (classic) SQL Server scenarios retirement