Short answer
When moving on-premises SQL Server to Azure, you choose among three main targets: SQL Server on an Azure virtual machine (most control, least change), Azure SQL Managed Instance (near-full SQL Server compatibility as a managed service), and Azure SQL Database (managed service for single-application databases). Size from measured workload, not from server specifications; plan network, identity, security, backup and disaster recovery before you move; and test with a pilot database before committing the estate.
Updated October 13, 2026. Written by Nathan Jaskot, founder of Everconnect Data Systems, a US-based, veteran-owned database consulting firm founded in 2010. Technical facts were checked against Microsoft documentation on that date (sources at the bottom).
At a glance
- VM: you manage the operating system and SQL Server; best for unusual needs or the quickest lift.
- Managed Instance: managed service closest to SQL Server; best for instance-level features.
- Azure SQL Database: managed service for single-application databases.
- Always plan: connectivity, identity, security, backup, high availability, disaster recovery, monitoring and cost controls.
Target options compared
| Target | You manage | Good fit | Consider carefully |
|---|---|---|---|
| SQL Server on an Azure virtual machine | Operating system, SQL Server patching, backups, high availability, tuning. | Full control, third-party agents, unusual configuration, fastest lift with the least application change. | More operational work stays with you. |
| Azure SQL Managed Instance | Databases, security configuration, performance. Microsoft handles patching and upgrades. | Legacy applications using SQL Agent, cross-database queries and other instance-level features. | Feature differences and limits; plan networking and migration method. |
| Azure SQL Database | Databases and application access. | One application, one database, modern patterns, elastic scaling needs. | No instance-level features; cross-database dependencies need design work. |
Sizing: measure, then add headroom
- Collect real metrics over a representative period that includes your busiest times (month-end, batch windows, reporting).
- Record the workload shape: CPU, memory, I/O operations and throughput, storage latency, tempdb use, log generation, connection counts and growth.
- Map to a size using Microsoft’s current sizing and assessment tools, then add headroom for growth and testing.
- Pilot the busiest or most complex database on the proposed size and compare with your baseline using Query Store.
- Right-size after cutover using Azure monitoring data, then revisit quarterly; cloud sizes can change far more easily than hardware.
The architecture decisions that are easy to miss
- Network: VPN or ExpressRoute, private connectivity, DNS, firewall rules and latency to applications that stay on-premises.
- Identity: how users and applications authenticate, including Microsoft Entra ID options and service accounts.
- Security: encryption at rest and in transit, key management, auditing, least-privilege access, threat detection.
- Backup and recovery: retention, restore tests and point-in-time recovery expectations; see how to prove backups and DR work.
- High availability and disaster recovery: zone and region choices against your recovery time and recovery point objectives.
- Licensing and cost: use existing licenses with Software Assurance through Azure Hybrid Benefit where eligible, reserved capacity for steady workloads, and cost alerts from day one.
- Operations: monitoring, alerting, patch windows (VM) and who responds at 2 a.m.
How to move the data
Method depends on the target and your downtime limit: backup and restore, the Managed Instance link or Log Replay Service for Managed Instance, availability-group based approaches for VMs, or Azure Database Migration Service. See our comparison of minimal-downtime methods, the earlier post on choosing between Azure SQL Database, Managed Instance and a VM, and our lessons from migrating legacy databases to Azure SQL Managed Instance.
Who can help
Everconnect Data Systems helps mid-sized and enterprise organizations assess their SQL Server environment, choose an Azure architecture and handle the migration. See our Microsoft Azure services and database migration services, or begin with the free SQL Server Migration & Upgrade Assessment (a self-assessment; it does not connect to your databases).
Why companies move a database to the cloud
- Cost model: pay for capacity as you use it instead of buying for the peak. This helps when demand varies; it does not always lower the bill for a steady workload.
- Scalability and availability: add resources or a standby copy without buying hardware.
- Less maintenance: with platform services, the provider handles hardware, patching and many backups.
- Disaster recovery and remote access: replicas in another region and access from anywhere, with the right controls.
Hybrid: keeping part on-premises
Many companies keep some databases on-premises and move others. That can be a sensible end state, not just a step along the way. Before choosing it, check:
- Infrastructure: what you already own and when it needs replacing.
- Cost: what you will pay for connectivity, licences and running two environments.
- The right mix: which workloads gain from the cloud and which have a reason to stay (latency, regulation, licensing).
- Integration and security: data that spans both sides needs reliable pipelines, consistent access control and one monitoring view.
Frequently asked questions
SQL Server on an Azure virtual machine, Azure SQL Managed Instance, or Azure SQL Database. The VM gives the most control, Managed Instance offers near-full SQL Server compatibility as a managed service, and SQL Database suits single-application databases.
Measure CPU, memory, I/O, throughput, latency, storage growth and peak patterns over a representative period, map them to a size with Microsoft's sizing tools, add headroom, pilot, then right-size after cutover.
SQL Server on an Azure virtual machine usually needs the least change because it is SQL Server running on a VM. Azure SQL Managed Instance is the managed option closest to SQL Server.
Eligible licenses with Software Assurance can be used through Azure Hybrid Benefit for supported Azure SQL options. Check the current terms with Microsoft or your licensing provider.
A secure connection between your network and Azure, such as a VPN or ExpressRoute, plus DNS and firewall planning. Requirements differ by target and migration method.
Right-size from measured usage, use reserved capacity for steady workloads, apply Azure Hybrid Benefit where eligible, set cost alerts and review usage regularly.
Yes. Migrate the most complex or busiest database first, compare performance with your baseline, and use what you learn to adjust sizing and the method for the rest.
Managed services reduce infrastructure work, but databases still need design, tuning, security, backup validation, cost management and incident response.
Yes. Everconnect Data Systems is a US-based, veteran-owned database consulting firm that helps organizations choose an Azure SQL architecture, plan the move and carry it out.
Related guides
- Upgrade SQL Server or move to Azure? A decision guide
- SQL Server migration with minimal downtime: methods compared
- SQL Server 2016 end of support: what to do now
- How to prove SQL Server backups and disaster recovery work
- How to choose between Azure SQL Database, SQL Managed Instance and SQL VM
Sources (checked October 13, 2026)
- Compare Azure SQL deployment options
- Azure SQL Managed Instance overview
- SQL Server on Azure Virtual Machines overview
- Managed Instance link overview
- Azure Database Migration Service overview
- Microsoft data migration documentation hub
About the author. Nathan Jaskot is the founder of Everconnect Data Systems, Inc., a US-based, veteran-owned database and technology consulting company he started in 2010. Everconnect works with SQL Server, PostgreSQL, MySQL, Oracle, Azure and AWS. Learn more about Everconnect.
Originally published October 13, 2026. Last updated November 9, 2026.







