Data Migration from On-Premise to Azure Cloud: A 5-Step Guide

Step 2: Choose the Azure target for each data set

The target depends on how much you want to change. Moving a SQL Server database to a VM is fastest; moving it to a managed service removes patching and backup work.

For SQL Server, Microsoft's SQL Server to Azure SQL migration overview compares the options. For the wider choice between rehosting, replatforming and refactoring whole applications, see our Azure cloud migration strategy guide.

Step 3: Pick the transfer method

Bandwidth, data size and how much downtime you can accept decide the method. Work out how long the initial copy takes over your actual uplink before choosing.

Azure Data Box is a physical device Microsoft ships to your data centre. According to the Data Box documentation, current devices have a maximum usable capacity of 120 TB or 525 TB. Data Box Heavy has been retired, and the 80 TB device is being retired region by region as the new devices roll out. A common pattern is to seed the bulk of the data with Data Box, then sync the changes online before cutover.

Is your migration toolchain up to date?

Many guides still describe tools Microsoft has retired or renamed. Check these before you write the plan.

• Data Migration Assistant (DMA): retired from 16 July 2025. Use the assessment features in Azure Migrate and the current Azure SQL migration guidance instead.
• Azure Active Directory: now Microsoft Entra ID. Plan identity for migrated data in Entra, including groups and role assignments for storage and databases.
• Azure Data Box 80 TB and Data Box Heavy: replaced by 120 TB and 525 TB devices, as above.
• Azure Database Migration Service: check which source and target combinations the current version supports for your engine before you commit to it.

Step 4: Migrate in tested waves

Move data in waves ordered by dependency and risk, and rehearse every wave before the real run.

1. Back up and test restores for every system in the wave. A backup you have not restored is a hope, not a plan.
2. Run a pilot on a representative but low-risk data set. Measure throughput and fix issues.
3. Seed the bulk copy online or with Data Box, while the source stays live.
4. Sync changes continuously, through database replication or repeated delta copies for files.
5. Freeze and cut over in an agreed window: stop writes, run the final sync, switch connection strings and DNS.
6. Keep the source read-only until validation and business sign-off, so you can roll back.

Run order usually follows dependencies: shared reference data first, then databases, then the applications and integrations that use them.

Step 5: Validate before you switch off the source

Validation proves that the data arrived complete and that applications behave as before.

• Completeness: row counts per table, file counts and total sizes.
• Integrity: checksums or hash comparisons on samples or full data sets.
• Function: application smoke tests, reports and integrations run by business users.
• Performance: query times and throughput against a baseline taken before migration.
• Security: access rights, encryption and logging confirmed in the new environment.

Set go/no-go criteria in advance and get sign-off from data owners. Decommission the source only after the agreed retention period.

Security and compliance during the move

Treat the migration window as a period of higher risk, because data is copied, staged and moved.

• Use private connectivity (VPN or ExpressRoute) or private endpoints rather than public endpoints for transfers.
• Encrypt data in transit and at rest; Data Box devices encrypt data on the device.
• Control access with Microsoft Entra ID roles and least privilege. Store keys and secrets in Azure Key Vault.
• Apply Azure Policy for region, encryption and tagging rules from day one, so non-compliant resources are blocked or flagged.
• Record where personal data is stored and processed, and update your GDPR records of processing.

The Cloud Adoption Framework migration guidance describes how to prepare landing zones and governance before data arrives.

After the migration: cost and performance

• Right-size databases and storage based on measured use, not on-premise allocations.
• Tier storage: move rarely used data to cool, cold or archive tiers with lifecycle rules.
• Commit where stable: reservations or savings plans for steady workloads; Azure Hybrid Benefit if you have eligible SQL Server or Windows Server licences.
• Monitor: Azure Monitor alerts for latency, errors, capacity and backup status.

How Opsio helps

Opsio plans and runs data migrations to Azure: discovery and assessment with Azure Migrate, target design, transfer method, wave planning, cutover and validation. Our Azure migration services cover the full project, and our database migration service handles SQL Server, Oracle, MySQL and PostgreSQL moves. After cutover, Azure managed services take over monitoring, backup and cost control.

Frequently asked questions

How do I migrate data from on-premise to Azure?

Inventory and classify the data, choose an Azure target for each data set, pick a transfer method based on size and bandwidth, migrate in tested waves with a final sync and cutover, then validate completeness, integrity and performance before switching off the source.

Which tool should I use to migrate SQL Server databases to Azure?

Use Azure Migrate for discovery and assessment. For the move, use Azure Database Migration Service for supported online migrations, or native backup and restore, log shipping or replication for SQL Server on VMs and Managed Instance. Data Migration Assistant was retired in July 2025.

When should I use Azure Data Box instead of an online transfer?

When the initial copy would take too long over your network or would saturate the link for production traffic. Data Box ships a device with 120 TB or 525 TB usable capacity; you copy the data locally, ship it back and Microsoft uploads it to your storage account.

How long does a data migration to Azure take?

It depends on data volume, bandwidth, change rate, the number of systems and how much testing you need. Calculate the transfer time over your real uplink, add rehearsal and validation time per wave, and plan cutover windows with the business.

Can a provider move our on-premise workloads into Azure securely?

Yes, if the plan includes private connectivity, encryption in transit and at rest, least-privilege access through Microsoft Entra ID, policy guardrails and logging. Ask the provider to show how each is configured and who holds administrative access during and after the project.