Why
Modernisation delivers step-change savings (30–70%) that incremental rightsizing cannot achieve. Serverless eliminates idle cost entirely for bursty workloads. Containerisation enables bin-packing and auto-scaling. Managed services eliminate operational overhead. Workloads running on migration-era architecture are permanently over-provisioned.
What
Systematically evaluate workloads for modernisation — containerisation, serverless migration, managed service adoption, or storage tiering — with a documented cost-benefit analysis before commitment.
How
Assess Modernisation Readiness per Workload
Use the architecture review findings (S7-01) to shortlist candidates. For each, assess:
| Factor | Assessment |
|---|---|
| Current architecture | VMs, containers, serverless, managed? |
| Traffic pattern | Steady-state, bursty, batch, event-driven? |
| State management | Stateless (easy to modernise) or stateful (harder)? |
| Team capability | Does the team have container/serverless experience? |
| Business criticality | Can we tolerate the migration risk? |
Build Cost-Benefit Analysis
For each candidate: estimate current monthly cost, projected cost on target architecture, migration effort (weeks and $), payback period, and risk assessment.
| Modernisation Path | Typical Savings | Best For |
|---|---|---|
| VMs → Containers (EKS/AKS/GKE) | 30–50% | Steady-state, multi-service |
| VMs → Serverless (Lambda/Functions) | 50–80% | Bursty, event-driven |
| Self-hosted DB → Managed | 20–40% | Operational overhead reduction |
| Monolith → Microservices | Variable | Independent scaling needed |
Prioritise and Execute
Rank candidates by payback period (shortest first). Start with the easiest modernisation that delivers meaningful savings. Document the pattern so subsequent workloads follow the same playbook.
Deliverable Checklist
- Modernisation candidates shortlisted from architecture reviews
- Cost-benefit analysis documented per candidate
- Priority ranking by payback period
- First modernisation executed (or approved for execution)
- Pattern documented for repeat use