5.1.3.1. Design for Azure Migrate
5.1.3.1. Design for Azure Migrate
💡 First Principle: A centralized migration hub that provides comprehensive discovery, assessment, and migration tooling is essential for simplifying complex migration projects and accelerating the move to the cloud with confidence.
Scenario: You are leading a project to migrate a large on-premises data center to Azure. This involves discovering all existing servers and applications, assessing their compatibility and costs for migration to Azure VMs, and then executing the migration of both Windows and Linux servers.
Azure Migrate is a free service that provides a centralized hub to discover, assess, and migrate on-premises servers, infrastructure, applications, and data to Azure.
Key Design Considerations:
- Discovery and Assessment:
- Agent-based/Agentless: Deploy lightweight appliances to discover servers, databases, and applications.
- Readiness Assessment: Analyze compatibility, performance, and cost estimates for migration to Azure VMs, Azure SQL Database, etc. Provides recommendations for right-sizing.
- Dependency Mapping: Identify dependencies between applications and servers to plan migration waves effectively.
- Migration Tools:
- Server Migration: Migrate physical servers, VMware VMs, Hyper-V VMs to Azure VMs.
- Database Migration: Migrate SQL Server and open-source databases to Azure SQL Database.
- Web App Migration: Migrate ASP.NET and Java web apps to Azure App Service.
- Data Box Integration: For large data transfers, integrate with Azure Data Box.
- Migration Strategy:
- Phased Migration: Migrate workloads in waves to minimize risk.
- Pilot Migrations: Conduct small-scale pilot migrations to validate the process.
- Cost Optimization: Leverage assessment reports to right-size Azure resources and optimize costs post-migration.
Design decisions in practice — the assessment settings ARE the assessment:
| Setting | Options | Why it matters |
|---|---|---|
| Sizing criteria | As-on-premises vs performance-based | As-on-premises copies the estate's existing over-provisioning into Azure and inflates the projection. Performance-based sizes from collected CPU, memory, disk and network utilization |
| Performance history | Day, week, month | Longer windows catch month-end and seasonal peaks. A one-day window on a workload with a monthly cycle will under-size it |
| Percentile | 95th, 99th | The 95th percentile ignores brief spikes. Use the 99th where those spikes are the workload |
| Comfort factor | Multiplier on the result | A buffer for growth. Raising it reintroduces over-provisioning, so keep it modest and justify it |
| Confidence rating | 1 to 5 stars | Reflects how many data points the appliance actually collected. A low rating means the assessment ran too briefly to trust |
Dependency analysis is the other half and answers a different question. Sizing tells you how big each server should be; dependency analysis tells you which servers must move together. It maps observed network connections, so it finds the integrations nobody documented — and moving one half of a chatty pair leaves its traffic crossing back to the datacenter over the WAN.
⚠️ Exam Trap: a low confidence rating is not a reason to raise the comfort factor. It is a reason to collect for longer — padding a bad estimate does not make it a good one.
⚠️ Common Pitfall: Ignoring the dependency analysis feature. Migrating a server without understanding its dependencies on other systems is a common cause of post-migration application failures.
Key Trade-Offs:
- Depth of Assessment vs. Speed: A thorough assessment takes more time upfront but significantly de-risks the migration and improves cost forecasting. A rushed assessment can lead to unexpected compatibility issues or cost overruns.
Reflection Question: How does designing for Azure Migrate, utilizing its discovery, assessment, and migration tools for diverse on-premises workloads, fundamentally provide a centralized hub that streamlines the end-to-end migration process, accelerating the move to the cloud with minimal disruption?
