"Can we just move it to the cloud as it is?" We hear that question in almost every first meeting about an old system. Usually it's the quoting tool or the customer database: still on Windows Server 2012, built on .NET Framework 4.x, and understood by two people who are booked out until March.
The board wants it in the cloud, the vendor wants a licence renewal, and engineering wants a rewrite. Each of them is right about something. Legacy system modernisation isn't one big decision, it's a series of small ones, made system by system, about how much to change and when.
Get the call wrong and you pay either way. Move a fragile system untouched and you carry its outages into a new monthly bill. Rebuild a stable one and you can lose a year recreating features nobody asked for. Below is how we help leadership teams choose between rehost, replatform and rebuild, and how to explain that choice in terms of cost, risk and time.
A shared vocabulary, finally
The big cloud providers now describe these options in nearly the same words. AWS lists seven migration strategies, the "7 Rs": retire, retain, rehost, relocate, repurchase, replatform, and refactor or re-architect. Microsoft's Cloud Adoption Framework uses a similar list and treats rebuild and replace as separate options, each tied to a business driver.
That helps when you're explaining a plan to a board. The pressure to actually decide comes from elsewhere, though. Unsupported operating systems and runtimes stop getting security patches. AI features need clean data and a usable API, which a 15-year-old monolith rarely offers. And the cost of keeping old platforms alive tends to creep up every year. So the question has shifted from "should we modernise?" to "which systems get which treatment, and in what order?"
Three paths, without the jargon
Be precise about what each one actually changes, because the labels get used loosely.
- Rehost (lift and shift). The application moves to cloud virtual machines unchanged. AWS is candid that this lets you scale without implementing any cloud optimisations that could save you time or money, though it adds that applications are easier to optimise once they're already in the cloud.
- Replatform (lift, tinker and shift). The application moves and some foundations get swapped for managed services. A self-managed SQL Server becomes Amazon RDS for SQL Server or Azure SQL Database; a VM workload gets packaged into containers. Code changes are small. The savings on patching, backups and licences often aren't.
- Rebuild (refactor, re-architect or rebuild). Significant parts get redesigned and rewritten for cloud-native services. AWS calls refactoring the most complex and costly of the migration strategies and, for large migrations, recommends modernising after the move wherever possible.
Two more options deserve a line on every shortlist. Retire whatever no longer earns its keep. Replace custom software with SaaS when you'd barely customise it, which is common for CRM and HR systems.
Legacy system modernisation trade-offs at a glance
Treat this table as directional, not a quote. Real cost and duration depend on system size, data volume, integrations and how much time your team can actually give the work.
| Factor | Rehost | Replatform | Rebuild |
|---|---|---|---|
| Upfront cost | Lowest | Moderate | Highest |
| Time to first value | Shortest | Short to medium | Longest |
| Delivery risk | Low, if the system is stable | Low to moderate | High without incremental delivery |
| Ongoing run cost | Often similar to today | Usually lower (managed services, fewer licences) | Potentially lowest, if designed well |
| Technical debt | Carried across unchanged | Partly reduced | Can be largely removed |
| Skills needed | Infrastructure and migration | Infrastructure plus some application work | A full product engineering team |
| Best fit | Stable systems with a short remaining life | Stable systems with obvious managed-service equivalents | Systems that block the business or can't be maintained |
Rehost: fast, but only for healthy systems
Rehost works for systems that run fine, aren't changing much and need to leave a data centre or a dying server quickly. Microsoft's guidance gives two tests we use verbatim. First, don't rehost problematic workloads, because moving them doesn't fix performance, reliability or architecture problems. Second, confirm the workload won't need modernising within two years. Fail either test and you'll probably pay twice.
A gotcha we see often: the rehosted VM keeps the same oversized spec it had on premises, runs 24/7, and the first cloud invoice becomes the argument against the whole programme. Right-size before cut-over, not after.
Replatform: where most business systems should land
This is the path we recommend most often. It goes after the parts of the stack that cost the most to run (the database, OS patching, licences) without reopening the business logic. Microsoft's indicators line up: simpler reliability and disaster recovery, less OS and licensing overhead, and an app the team can containerise with moderate effort.
There's a second benefit people miss. Replatforming forces you to build the deployment pipeline and cloud skills you'd need for anything bigger later. It's a good rehearsal.
Rebuild: when the system is the bottleneck
A rebuild is justified when the current system actively gets in the way. Releases are slow and scary, test coverage is close to zero, nobody can safely change the code, or the architecture can't meet a new requirement. AWS lists these exact signals, including a monolith that is already hindering efforts to deliver product quickly and applications that are difficult to test.
Even then, skip the big-bang rewrite. Martin Fowler's Strangler Fig pattern builds new capability alongside the old system and moves behaviour across a slice at a time, usually behind a routing layer such as an API gateway or reverse proxy. Value arrives early and you can stop when it's good enough. If the rebuild means breaking up a monolith, read our take on when monolith to microservices pays off before you commit to an architecture.
The per-system checklist
Run each system through these questions in a workshop with the business owner in the room. The answers usually point to one path.
- Does it still create business value? If not, retire it and stop paying for it.
- Is there a mature SaaS product that does the job with little customisation? Then look at replace before spending on custom work.
- Will its scope or architecture need to change in the next two years or so? If not, rehost or replatform. If yes, replatform now and rebuild in stages, or rebuild directly.
- Where does the run cost actually sit? If it's mostly databases, licences and patching, replatforming those pieces usually pays back fastest.
- Is it stable and understood? Stable systems move well. Fragile ones need fixing first, not moving.
- Can the team deploy it repeatably? If deployment is a person following a Word document, introduce infrastructure as code and a pipeline first. Every path depends on it.
- What happens if it's down for a day? High-impact systems need recovery targets, backups and failover designed in, whatever the path (see our guide to cloud resilience).
- Do we have the skills and the time? Microsoft's advice is blunt: modernise when the team has both, and defer when it doesn't.
How JRC Software runs a modernisation assessment
We begin with an inventory, not slides. Our engineers in Melbourne and Sri Lanka list every system in scope, pull the real hosting and licence costs, map integrations (including the nightly CSV export nobody remembers until it breaks), and check OS, runtime and database versions against vendor support dates.
Over the following weeks we score each system against the checklist above with your business owners. What you get at the end is a one-page portfolio view: retire, rehost, replatform or rebuild for each system, in a suggested order, with indicative effort and the risks we found. You can take it to a board or to another supplier.
We'll also tell you what not to do. Typically that's: don't start a full rewrite before you have automated deployment, don't rehost a system that's already falling over, and don't sign a multi-year licence renewal until the assessment is done. When we do the delivery, it goes in that order too: pipeline and infrastructure as code first, replatform the expensive parts next, then strangler-style slices for the pieces that genuinely need rebuilding. The old system keeps running until each slice is proven.
The engagement model is outlined on our cloud engineering and modernisation service page.
Not sure which path each of your systems should take? Book a modernisation assessment and we'll give you a per-system recommendation with the cost, risk and timeline trade-offs written down.
Frequently asked questions
What is the difference between rehost and replatform?
Rehost moves an application to the cloud unchanged, so it runs on cloud virtual machines much as it did on premises. Replatform also moves it, but swaps some underlying components for managed services, such as a managed database or containers, with minimal code changes. Replatforming takes a bit longer but usually cuts ongoing run and maintenance costs.
Is rebuilding a legacy system ever cheaper than rehosting it?
Rarely upfront, because rebuilding is the most complex and costly path. Over the system's life it can work out cheaper if the old platform carries heavy licensing, maintenance or support costs, or if it's slowing product delivery. It comes down to how long the system needs to last and how much it will change, which is what a per-system assessment puts on paper.
How long does legacy system modernisation take?
It depends on the path. A rehost of a stable application can be quick once networking and landing zones are ready, while rebuilding a core platform is usually a staged effort over many months. We plan in increments that each deliver something useful, rather than one long programme with a single go-live date.
Should we modernise during migration or after?
For large migrations, AWS recommends rehosting, relocating or replatforming first and modernising afterwards, because refactoring mid-migration is hard to manage at scale. Microsoft suggests modernising during migration when the team has the skills, the time and a clear business case. Many organisations do both: move most systems quickly and modernise a few high-value ones in parallel.
Cloud Architecture
Book a modernisation assessment
Tell us where you're stuck and we'll come back with practical next steps.
Get in touch