Four layers to review
| Layer | What to inspect |
|---|---|
| Compute | CPU allocation, memory headroom, storage behavior and competing workloads. |
| Network | Actual broker or service endpoints, routing tests and administration paths. |
| Access | Account permissions, Remote Desktop exposure and credential handling. |
| Recovery | Independent backups, application startup order and restoration evidence. |
A server can pass one layer and fail another. For example, ample CPU does not prove a stable broker connection, and reachable Remote Desktop does not prove that an EA is initialized correctly. Define checks for the workflow you actually depend on.
Keep claims configuration specific
Processor model, disk arrangement, data-center placement, network protection, backup service and support commitments belong in the written specification. A region label or generic illustration is not evidence of a particular facility or measured performance.
For network-sensitive work, ask for a test path to the destination. For capacity-sensitive work, provide a representative resource sample. For recovery-sensitive work, agree what is restored, from which copy and by whom.
Separate maintenance from incidents
- Maintenance: planned operating-system updates, application changes, resource upgrades and restart tests.
- Monitoring: checks for host health, application process state, connection behavior and backup completion.
- Incident response: a contact route, responsibility boundary and information required to investigate.
- Recovery: a tested restoration process for applications, settings and necessary data.
Do not assume that all four are included in a server price. Ask which tasks are supplied and which remain with your own administrator or another application provider.
Use a practical acceptance checklist
- Confirm the delivered specification against the quotation.
- Secure access and establish the intended administration process.
- Run the representative application workload and review resource and connection data.
- Perform a planned restart and a restoration test.
- Record operational ownership, maintenance arrangements and escalation contacts.