Cloud Migration for the Midmarket — Step by Step, Not Big Bang
Source: Dev.to
The Lift-and-Shift Trap
Most cloud migrations in the midmarket fail not because of technology — they fail because of strategy.
The classic mistake: an IT services company promises to “move everything to the cloud,” lifts VMs from an on‑premise data center into AWS or Azure, and suddenly the company pays triple for the same performance. No auto‑scaling, no automation, no modernization. Just a more expensive version of the status quo.
That’s not a cloud project. It’s a relocation with the wrong address.
What a Real Migration Requires
Before a single server moves, three questions need answers:
1. What do we migrate — and in what order?
- Not everything belongs in the cloud.
Example: An Oracle database with 15 years of stored procedures? Probably not first in line.
Example: A stateless API service with clean interfaces? Perfect starting point. - Sequencing decides whether you succeed or burn out.
We use a simple assessment matrix:
| Factor | Guidance |
|---|---|
| Dependency complexity | High → migrate later |
| Business criticality | Critical → more testing, not first wave |
| Cloud readiness | Stateless, containerizable, APIs available |
| Cost‑benefit ratio | Prioritize where cloud saves the most |
2. Which migration pattern fits?
There’s no one‑size‑fits‑all. Common patterns from the AWS 7‑R model:
- Rehost (Lift & Shift) – Move the VM, done. Fast, but no cloud advantage.
- Replatform – Small adjustments (e.g., managed database instead of self‑hosted).
- Refactor – Rebuild cloud‑native (containers, Kubernetes, serverless). More effort, maximum payoff.
- Retire – Shut it down. Some systems genuinely aren’t needed anymore.
- Retain – Deliberately keep on‑premise. A valid decision too.
For most midmarket companies, Replatform is the sweet spot: leverage managed services without rewriting everything.
3. How do we control risk and cost?
Cloud costs spiral when nobody’s watching. According to the Flexera 2025 State of the Cloud Report, companies overspend by an average of 28 %.
Recommendations
- FinOps from day one – Budgets, alerts, cost explorer. Not after the first invoice shock.
- Reserved Instances or Savings Plans for predictable workloads.
- Rightsizing before migration — don’t run the same oversized instance in the cloud.
- Kill switch — Every migrated service must be rollback‑capable (minimum 30 days).
Common Pitfalls
Networking Gets Underestimated
On‑premise applications often communicate over local networks with sub‑1 ms latency. In the cloud, that jumps to 5–20 ms. For some applications the difference is irrelevant; for others, performance collapses.
Solution: Perform a network assessment before migration. Identify latency‑sensitive workloads and plan accordingly (same availability zone, VPC peering, or deliberately keep on‑prem).
Compliance Gets Forgotten
Once data leaves your data center, new rules apply: GDPR‑compliant hosting location, data‑processing agreements, technical‑organizational measures. Regulated industries (finance, healthcare) have additional requirements.
Solution: Conduct a data‑protection impact assessment before migration. Choose cloud region deliberately (e.g., EU‑only). Enforce encryption at‑rest and in‑transit by default.
The Team Isn’t Ready
The best cloud architecture is worthless if the team can’t operate it. “We’re migrating to the cloud” without training is like “we’re switching to electric vehicles” without charging infrastructure.
Solution: Invest in cloud skills alongside the migration. Pursue AWS/Azure certifications, pair‑program with experienced cloud engineers, and create runbooks for day‑2 operations.
What It Actually Costs
The question every CEO asks: “What does this cost me?”
A realistic estimate for a mid‑sized company (50–500 employees):
| Phase | Duration | Investment |
|---|---|---|
| Assessment & Strategy | 2–4 weeks | Consulting + internal time |
| Pilot Migration (1–2 services) | 4–8 weeks | Engineering + cloud costs |
| Main Migration (waves) | 3–12 months | Engineering + cloud costs + parallel operation |
| Optimization & Stabilization | Ongoing | FinOps + monitoring |
When done right, cloud infrastructure costs typically land at 60–80 % of previous on‑premise costs. With a pure lift‑and‑shift, they can rise to 120–150 %.
Our Approach
We run cloud migrations for midmarket companies. No slide decks — hands‑on engineering:
- Assessment — What do you have, what belongs in the cloud, and in what order?
- Architecture — Target architecture with Kubernetes, managed services, infrastructure as code.
- Pilot — Migrate one service, learn, adjust.
- Waves — Migrate remaining services in planned waves.
- Enablement — Train your team to operate the cloud independently.
No vendor lock‑in, no black box. Everything transparent, everything documented, everything yours.