Cloud migration step by step: from assessment to operations

By Samir Hassen, Founder Published 6 min read

A cloud migration is rarely a single move. It is a series of decisions: which systems to move, in what order, in what way and what should stay on-premises. Migrations that go well usually have one thing in common. The decisions are made before the move starts, not during it.

This guide walks through the steps from the first assessment to operations. It builds on Microsoft’s Cloud Adoption Framework and on how we plan migrations to Azure ourselves.

Start with why

Before you look at the technology: what do you want to achieve? Common reasons are servers or platforms that have reached the end of their life, costs that need to become clearer, security that needs to improve or systems that need to scale. More and more companies also want a foundation for AI and automation.

The goals steer everything else. If the goal is to leave a data center by a certain date, speed matters most. If the goal is lower costs, moving the systems as they are is rarely enough. Write the goals down, together with who owns them. They become the yardstick for every decision later on.

Step 1: Map your environment

You cannot plan what you cannot see. Make an inventory of your systems with:

  • what each system does and who owns it in the business;
  • which servers, databases and licenses it uses;
  • which other systems it talks to, its dependencies;
  • requirements on availability, security and where data may be stored.

Dependencies are often where the surprises are. A system that looks standalone may read from a shared database or send files to another system every night. Move one without the other and something stops working.

Microsoft’s Azure Migrate tool helps with inventory, dependency analysis, assessing what fits in Azure and estimating costs, and it can also carry out the move itself. According to Microsoft, Azure Migrate is a free service, though partner tools may charge. The Azure resources you run during and after the move are billed as usual. The tool does not replace talking to the people who know the systems, but it gives you a solid base of facts.

Step 2: Choose a path for each system

Not every system should move the same way, and not every system should move at all. Microsoft describes eight paths:

  1. Retire. Systems that no longer add value are switched off instead of moved.
  2. Retain. Systems that work and have no reason to move stay where they are. Some have to stay, because of regulation, technical dependencies or business continuity.
  3. Rehost (often called lift and shift). The system moves with as few changes as possible. It is fast, but rarely brings the biggest gains.
  4. Replatform. The system moves to a more modern hosting environment, such as a managed database, with small code changes.
  5. Refactor. The code is improved internally, without new features, so it works better in the cloud.
  6. Rearchitect. The system’s design is reworked to scale better and be easier to develop.
  7. Rebuild. The system is developed anew with cloud-native services.
  8. Replace. A ready-made SaaS product takes over instead of custom development.

Ending up with a mix is normal. Some systems are rehosted to get started, some are modernized and some stay on-premises. The result is often a hybrid environment, and that is perfectly fine as long as it is designed on purpose.

Step 3: Build the foundation

Before the first system moves, you need somewhere to move it to: a landing zone. It is the foundation in Azure with structure, networking, access, security, monitoring and cost tracking. Read more in our guide What is an Azure landing zone?

The foundation does not have to be complete on day one. But what is hard to change later, such as the network’s address plan and how access is granted, should be decided.

Step 4: Plan in waves

Large migrations are divided into waves: groups of systems that move together. Systems that depend on each other go in the same wave.

Some principles from Microsoft’s framework and from practice:

  • Start with simpler, lower-risk systems. The first wave is also a test of the process itself.
  • Move test environments before production. You find the problems where they cost least.
  • Plan only the next wave in detail. Each wave teaches you something that affects the next.
  • Decide how much downtime each system can take. Some can move during a planned maintenance window. Others need replication so the interruption is very short.
  • Have a plan to roll back. Decide in advance what happens if something goes wrong, and who makes the call.

Step 5: Move and test

The move itself is often the shortest part of the work, if the preparation is done. Data is synchronized, the system is started in Azure and tested before traffic is switched over:

  • does the system work as it should, including with the systems it depends on?
  • is performance good enough under normal load and at peaks?
  • are access rights, logging and backup in place?

Tell users in good time what will happen and when, and have someone available to answer questions after the switch.

Step 6: Optimize and decommission

Once the system runs in Azure, the next phase begins. Follow up how resources are used and adjust their size. Switch off what is not used, automate environments and track costs continuously with clear owners.

Do not forget the old environment. Servers, licenses and contracts that are no longer needed should be decommissioned, or you pay for both environments. This is also where the gains from the move become visible.

Common pitfalls

  • Dependencies discovered too late. A common cause of outages during a move.
  • Everything rehosted, with costs expected to fall. Without adapting the systems, the cloud can cost more than the old environment.
  • Security added afterwards. Access and networking are much easier to get right from the start.
  • Nobody owns the costs. Without tagging and owners, the bill grows without anyone noticing why.
  • Everything at once. A big move over one weekend is hard to test and hard to roll back.

Frequently asked questions about cloud migration

How long does a cloud migration take?

It depends on how many systems need to move, how they connect and your requirements. One well-defined system can be quick. A whole environment moves in several waves. After an assessment, a realistic timeline can be set.

Do we have to move everything?

No. Some systems should stay on-premises, for example because of requirements on where data is stored or dependencies on local equipment. Others can be retired or replaced. The goal is the right environment for each system, not having everything in the cloud.

Can we move without downtime?

For many systems the interruption can be kept very short, with replication and a well-prepared switchover. For others, a planned maintenance window is simpler and safer. It is decided per system, during planning.

Start with an assessment

The best first step is to understand what you have and what you want to achieve. Read more about how we work with cloud architecture and migration to Azure, from assessment to operations.

Want to talk about your environment? Book a call with our founder, and we will go through where you stand and which first step fits.

Samir Hassen

Samir Hassen is the Founder of Son of a Coder. Leads business strategy, solution architecture and client communication.

More articles about cloud architecture

Contact me