Back to Article List

What actually moves when you move to the cloud

A technician crouched beside an open server cabinet, tracing a bundle of cables while taking stock of what is in the rack.

The pressure to move a business into the cloud usually arrives as a single question, put either by a vendor or by a server that has started making a noise: are you moving or aren't you? Firms that come through a migration without a bad quarter rarely answer that question, because it turns out not to be the useful one. The decision that matters isn't about the business at all. It's about each application the business runs, taken one at a time. We'll look at how that call gets made, why the two options that involve moving nothing are often worth the most, and what to check about the way out before you commit to a way in.

The decision isn't about your business, it's about each application

Amazon Web Services publishes the list that most migration plans quietly borrow from, and it has seven entries rather than two. Its prescriptive guidance calls them the 7 Rs: rehost, relocate, replatform, refactor, repurchase, retire and retain. Rehosting means moving a server largely as it stands. Refactoring means rebuilding the thing so it works the way cloud services expect — expensive, and occasionally worth it. Five other answers sit between those two, and a migration plan is mostly the work of deciding which one applies to which piece of what you already run.

That is more work than a yes or a no. It is also the part that decides the bill.

The work starts with an inventory — the hardware, the applications, the databases, the network gear, and whatever is running on the machine under someone's desk. For a dental practice, a two-branch insurance agency, or a small engineering firm, that list is usually shorter than anyone fears and stranger than anyone expects. Three or four entries on it will be things nobody has thought about in years. Those are the interesting ones.

Two of the seven options involve moving nothing at all

Retire is the strategy for the applications you plan to decommission or archive outright. Every business accumulates them, and they are easy to name once you go looking: the reporting tool bought for a project that finished, the second customer database inherited in a merger, the file share holding a decade of superseded quotes. Retiring those costs nothing to run. Each one also drops out of every later decision, which is what makes retire the only entry on the list that shrinks the project rather than growing it.

Retain covers the applications that stay where they are, either because they should or because you aren't ready to move them yet. A workstation wired to a cutting machine on the shop floor qualifies. So does a practice management system whose vendor charges noticeably more for the hosted version, and a database a regulator expects to sit in a particular place. Choosing retain isn't a failure of nerve. It's a decision with a date on it, and it gets revisited when the reason behind it changes.

The date that usually forces the question

Most small businesses don't start this exercise because they read something about the cloud. They start it because a support date is coming up. Extended support for Windows Server 2016 ends on 12 January 2027 on Microsoft's own lifecycle listing, which is a little over five months away. End of support doesn't mean the server stops working. It means Microsoft stops issuing security updates for it, so anything found in it after that date stays open for as long as the machine is plugged in.

A deadline like that is a good reason to do the inventory and a poor reason to skip it. The tempting move is to lift everything onto rented servers before the date and sort out the details later — which turns a hardware problem into a monthly bill and leaves the retire-and-retain work undone. That work doesn't disappear. It just costs more to do afterwards.

Check the way out before you commit to the way in

Egress fees are what a provider charges to send your own data back out of its network, and for years they were the quiet reason a migration was easier to start than to reverse. That has changed, at least for the raw data. Google Cloud dropped transfer charges for customers leaving its platform in January 2024 and applied it to all customers globally, and Amazon Web Services followed two months later with free data transfer out for customers moving to another provider or back on-premises. The Amazon version runs through its support team rather than automatically, and as of September 2025 it allows 90 days to complete the move, which is generous for a file server and tight for a busy database.

That is a real improvement, and it solves the smaller half of the problem. Moving the bytes was never the expensive part. What holds a business in place is the reshaping done on the way in — the database converted to a managed service, the logins folded into a new identity provider, the integrations rewritten against a different set of interfaces. None of that comes back for free, which is a fair argument for rehosting first and refactoring later, once you know the thing actually works where you put it. It is also why the licensing upheaval that pushed so many teams off VMware cost some of them a weekend and others most of a year.

One step belongs before all of it, and it is the one most often assumed rather than checked. Every workload gets backed up and restored somewhere before anything moves. A backup nobody has restored is a belief rather than a backup, and the tools built into Windows are narrower than their names suggest.

If you'd like a second pair of eyes on the inventory

The shift here is a small one, and it holds up well beyond this project. A move to the cloud isn't a position a business takes. It's a set of decisions about specific things, most of which can be made independently, and several of which turn out to be leave it alone or switch it off. Businesses that treat it that way spend less and break less, mostly because they end up moving less.

We do this work for small and mid-size companies across a fair range of setups, from one ageing server in a back office to a rack that has grown by accretion. Some of it is managed hosting for what moves. If you'd like a second opinion on your own list, we're glad to look at it with you. For an accounting or tax firm the answer is rarely all-or-nothing, because prior years and the tax software have rules of their own.

Sources

About the migration strategies, AWS Prescriptive Guidance. Windows Server 2016, Microsoft Lifecycle. Eliminating data transfer fees when migrating off Google Cloud, Google Cloud, January 2024. Free data transfer out to internet when moving out of AWS, AWS News Blog, March 2024.