DFdom@fradley
← ~/writing
$ cat lift-and-shift-end-date.md

Lift-and-shift with a written end date

Whether to move a workload to Azure as it is or rebuild it first decides most of a migration's cost and most of its risk. It usually gets argued as taste. One person says a straight lift-and-shift buys your old problems back at cloud prices. Another says rewrites get funded once and delivered never. Both views come from real failures, and neither settles the decision.

The question can't be answered for an estate. It can be answered for each workload, and the variable that decides it is commercial rather than technical.

How long does this workload have left?

Is it going to matter for another eighteen months or another five, and will it change while it does?

A workload with eighteen months left and no roadmap is not a re-architecture candidate, however ugly it is. A workload the business plans to build on for five years is not a rehost candidate, however well it runs today. The technical assessment prices each option: how tangled the dependencies are, where the data lives, what breaks if latency doubles. It can't pick one; that depends on the commercial life, and the person who knows it is often not in the migration workshop.

This is why triage stalls in so many programmes. The technical assessment gets done thoroughly, then the decision is made on preference because the input that would have settled it never arrived.

Retire first

Every migration finds workloads nobody needs, and the cheapest workload to migrate is the one you switch off. Doing that pass first reduces every estimate that follows.

The obstacle isn't technical. Switching something off needs a named person to say out loud that they don't need it. To make that easier, stop the service, keep the backup, delete the backup after ninety days. If nobody has called by then, the workload was as unloved as it looked. If someone does, you can put it back the same afternoon. Repeat the pass when the migration is half done, because a second sweep finds more.

The four moves

Move Right when Watch for
Rehost The workload is stable and unloved with a known end date, or a datacentre exit clock is running and the clock outranks elegance No follow-up. See the next section
Replatform A managed service removes operational toil without a code change: SQL Server onto Azure SQL Managed Instance, a self-hosted broker onto Service Bus, a hand-patched web tier onto App Service Usually skipped, because rewrites are more interesting to talk about than swapping who patches the thing at 2am
Re-architect The workload is on a growth path, the business wants to change it anyway, or the scaling economics are already broken Re-architecting a stable, low-change workload is engineering as a hobby, funded by someone who thinks they are buying a business outcome
Retire Nobody needs it Nobody will say so without the backup safety net above

In most estates replatform gives the best return for the risk. The same judgement applies to the target. AKS earns its operational cost for services with real scaling demands and a team that owns them. A great deal of what gets pushed into clusters by default would be happier as Functions or managed PaaS, with nobody on call for a control plane.

Lift-and-shift needs an end date

Rehosting with no follow-up is the most reliable way for a cloud programme to lose its business case. The estate lands in Azure at roughly what it cost in the datacentre, sometimes more, because a VM sized against a five-year hardware refresh is the wrong shape for consumption billing. Everyone knew that at the time. The savings were always going to arrive in phase two, when the rehosted workloads were replatformed properly.

Phase two rarely gets scheduled. The programme that promised it closes, the plan stops being opened, and the people who understood the intent move on. What remains is a subscription full of VMs with no recorded future.

A date in a migration plan is hard to track. Put a review date and an owner on every rehosted workload as tags (the technical section at the end has the example), and "which rehosted workloads are past their date" becomes a query for the Monday meeting. Add a policy that audits virtual machines arriving without the tags, so undated workloads stop accumulating, and the information stays after the people who knew it have moved on.

Land it as code

Phase two usually fails for a duller reason than lost will. The rehosted workload arrived as a click, so every later change to it is priced at hand-built prices. Nobody is certain what it depends on, no change is reviewable, and the replatform gets quoted as a project with a discovery phase bolted to the front.

Bring the workload over as code and the same change is a diff. Swapping a VM's self-managed database for a managed instance becomes a pull request against a file that already describes the workload. The distance between rehost and replatform becomes an edit, which makes phase two far more likely to happen. So ask for Terraform from the migration, rather than a wave of portal work tracked in a spreadsheet.

Sequencing

Get the landing zone ready before volume. Migrating forty workloads into a governance model you are about to change means doing it twice, the second time under the pressure of a live estate. Settle policy, network, identity and tagging on the first few, then open the taps.

Order the waves by risk: lowest-risk stateless workloads first, to prove the pattern, then the harder ones.

What keeps this honest is an error budget. Define service level objectives (SLOs) on the critical user journeys and agree the budget policy with product leadership before the first incident rather than during it. When the budget is used up, migration slows and reliability work takes priority, as already agreed.

Every workload gets one of four answers, and every rehost gets a date.

For your platform team: the end-date tag

One tag set on every rehosted resource:

tags = {
  owner     = "payments"
  migration = "rehost"
  review_by = "2027-03-31"
}

The tag policy and the Resource Graph queries are in the blog-examples repo.

More useful tidbits coming, one a week.

discuss on linkedin → more writing