Cloud architecture review: reliability and cost together

A cloud architecture review connects expenditure to useful work and reliability to tested recovery.

A cloud architecture review connects expenditure to useful work and reliability to tested recovery. A bill can reveal an idle resource, but it cannot tell you whether removing it breaks a recovery plan. Review one business workload at a time and make the consequences of each proposed saving or resilience improvement explicit.

Inventory ownership and dependencies

Map accounts, regions, networks, data stores, compute, queues and third-party services. Identify the owner and environment of each important resource. Compare infrastructure definitions with deployed resources to find configuration drift. Include identity, DNS, certificates and deployment tooling in the map: an application can be healthy while a shared dependency prevents customers from reaching it.

Ask for demonstrated recovery

Agree the acceptable interruption and data-loss window for the workload. Inspect restoration evidence, backup isolation and the order in which dependencies return. Redundant application instances do not protect against every database or account-level failure. Trace a credible failure scenario and record the manual steps, credentials and people needed to restore service.

Normalise cost by useful activity

Separate recurring baseline cost from usage-sensitive cost and unusual events. Compare expenditure with completed work such as processed orders or active workspaces, using a consistent definition. Inspect storage retention, data transfer, idle environments and oversized resources. Before resizing, evaluate headroom, peaks and recovery requirements. A lower monthly total is not an improvement if it shifts cost into incidents or slower delivery.

Create a reversible change queue

For every recommendation, record current evidence, expected benefit, risk, owner and rollback condition. Apply configuration changes in a controlled environment and monitor both service quality and spend afterward. AWS Well-Architected provides a broad review structure; use provider guidance to supplement workload evidence, not as a substitute for it. Keep accepted exceptions visible with a review date.

• Related service →
• Code audit or penetration test: choose the evidence you need →
• Software architecture review: the evidence checklist →

• AWS Well-Architected — Reference documentation →