Why
Architecture decisions lock in cost profiles for years. A single review of a top-10 workload can identify savings that exceed months of ticket-level rightsizing. Incremental optimisation has diminishing returns — at some point, the architecture itself is the bottleneck.
What
Establish periodic architecture reviews with a cost-efficiency lens. Review top-spend workloads at least annually against Well-Architected cost optimisation pillars or equivalent frameworks.
How
Identify Review Candidates
Rank workloads by monthly spend. Select the top 5–10 for review. Prioritise workloads that have not been reviewed in >12 months, have growing unit costs (S9-01), or are running on migration-era architecture.
Conduct the Review
For each workload, a cloud architect spends 2–4 hours reviewing against a cost lens:
| Review Area | Questions to Ask |
|---|---|
| Compute model | Right instance family? Could this be containerised, serverless, or spot-eligible? |
| Scaling | Is auto-scaling configured and tuned? Are we over-provisioned for peak? |
| Storage | Are we using the right tier? Are lifecycle policies in place? |
| Data transfer | Are we paying for cross-AZ/cross-region unnecessarily? VPC endpoints vs NAT? |
| Database | Right engine? Right instance? Read replicas justified? Caching layer? |
| Architectural alternatives | Managed service vs self-hosted? Event-driven vs always-on? |
Document Findings and Create Action Items
For each finding: current architecture, proposed alternative, estimated savings ($/month), implementation effort (weeks), risk level. Create optimisation tickets for approved changes.
Establish Annual Cadence
Schedule annual reviews for top workloads. Add ad-hoc triggers: unit cost rising, budget variance >10%, or major scaling event.
Deliverable Checklist
- Top 5–10 workloads identified for review
- Architecture reviews completed (2–4 hours each)
- Findings documented with savings estimates
- Action items created as optimisation tickets
- Annual review cadence scheduled