Start a project

Moving a legacy system to the cloud without stopping the business

A step-by-step approach to moving an old business system to the cloud while staff keep working: assess, migrate in stages, cut over, keep a way back.

Many businesses in the Middle East run on a system that was installed years ago and never left the building. It sits on a server in a back room. The person who wrote it has moved on. It works, mostly, and everyone is afraid to touch it. Then a power cut, a failed disk or a security scare makes the risk impossible to ignore.

Moving that system to the cloud is possible without closing the business for a week. It takes patience and an order of work.

The first stage is an inventory, and it always turns up surprises. Write down:

  • Every application and database, and which version each runs on.
  • Every connection to something else: the accounting package, a payment terminal, a label printer, a spreadsheet someone imports every morning.
  • Scheduled jobs that run at night and that nobody remembers setting up.
  • Who uses the system, from where, and at what hours.
  • Where the source code is, if it exists, and whether it matches what is running.
  • What the system depends on that is no longer supported by its maker.

Ask the people who use it daily. They know the workarounds, and the workarounds are part of the system.

Choose the right kind of move

Not every system should be moved the same way. There are four broad choices.

  • Move it as it is. The same software runs on a virtual machine in the cloud. This is the fastest route and solves the back-room risk, but the software is as old as before.
  • Move it and adjust. The application is changed just enough to use managed cloud services, such as a managed database with automatic backups.
  • Rebuild it. The system is rewritten on a modern stack. This is the most work, and the right answer when the old technology is at the end of its life or the business has outgrown the design.
  • Replace it. A ready-made product takes over, and the old data is imported.

Most real projects combine these. A common pattern is to move the system as it is first, to make it safe, and then rebuild it piece by piece.

Move in stages, not in one night

The big overnight switch is the riskiest plan there is. A safer approach is gradual replacement. The new system grows beside the old one and takes over one function at a time. Reports might move first, because they only read data. Then a single module, then the next. Users are sent to the old or the new system depending on what they are doing, often without noticing.

Move in stages, with the old system still running beside the new.
Move in stages, with the old system still running beside the new.

Each stage is small enough to test properly and small enough to undo. The business sees progress every few weeks, not a promise of a launch a year away.

Treat the data as the hard part

Code can be rewritten. Data cannot be invented. Old databases contain duplicates, empty required fields, dates in several formats and Arabic text stored in old encodings that display as question marks when moved carelessly. Plan for it:

  • Clean the data before the move, not after.
  • Run the migration as a rehearsal several times, on a copy, and compare totals with the original.
  • For a period, keep the old and new databases synchronised so both hold the same records.
  • Agree with the business on what history is moved and what is archived.

Plan the cutover and the way back

When the final switch comes, choose a quiet period. In much of the region that means the weekend, which falls on different days in different countries, and it means avoiding month-end and peak season. Announce a short freeze on changes. Take a final copy, switch, and have real users check real tasks before the doors open.

Once you are there: monitoring, backups and costs you can see.
Once you are there: monitoring, backups and costs you can see.

Decide in advance what would make you go back, and make sure you can. Keep the old system intact and untouched for an agreed period. A rollback plan that has never been tested is a hope, not a plan.

What changes once you are there

Running in the cloud is different from owning a server. Costs become monthly and vary with use, so someone needs to watch them. Backups, monitoring and security updates can be automated, but they must be set up and checked. Access should be tied to named accounts with multi-factor sign-in.

Choose the hosting region with care. Some sectors and countries have rules about where data is kept. Cloud providers, Microsoft Azure among them, operate data centres in the region, but the requirement itself should be confirmed with your regulator or legal adviser.

Connectivity matters too. If staff depend on the system all day, an office with one unreliable internet line needs a second.

How Angry Monkey Agency can help

Angry Monkey Agency moves and rebuilds business systems on .NET and Microsoft Azure, and provides managed Azure hosting once they are there. We work in stages so that staff keep working throughout. If you have a system everyone is afraid to touch, we can start with an assessment and a plan.

We host what we build on Coconut Sharp, our own platform for .NET. Angry Monkey Shield watches and hardens what runs in it. The same team has worked on campaigns and digital work for brands such as stc, American Express, Burger King and Nissan. Read how we work in software development, look at the work or start a project.

Chat with us on WhatsApp