Record ordinary operation and busy periods.
List the applications, users, scheduled jobs and external services. Capture CPU, memory, storage and network use over a period that includes representative peaks. Record symptoms as well as averages: slow checkouts, a growing queue or an overnight job that misses its deadline.
If the application is new, identify assumptions and prepare a realistic test workload. Include expected growth and the business consequence of interruption. This helps distinguish a small experimental server from a production service with a defined support and recovery need.
Ask what the allocation means.
Clarify virtual CPU allocation, memory availability, disk capacity and any separate storage performance limits. Ask how network port speed relates to permitted transfer and traffic policies. A resource label is useful only when its allocation and limits are understood.
Leave capacity for logs, updates, temporary files and backups. Check whether resizing needs a restart, data move or different service. For storage-heavy and memory-heavy applications, compare a balanced configuration with the resource that is currently limiting performance.
Choose the operating system and its owner.
Use application compatibility, licence terms and team capability to decide between Linux and Windows. Include required releases, libraries, databases and control panels. Check the maintenance lifecycle and the work needed to keep the environment supported.
Assign operating system updates, secure access, application deployment, monitoring and restoration. If management is required, ask for the task list and support boundaries. A control panel may make common tasks easier, but the remaining administration still needs an owner.
Compare complete proposals and test the result.
Normalise the quotations: include setup, licences, backup capacity, retention, transfer, management and migration. Check the billing period, tax treatment, location and contract terms. Record material differences rather than assuming a cheaper compute line means a lower overall operating cost.
Agree an acceptance checklist before ordering. It should cover important application transactions, administration, monitoring and a restore check. Keep the configuration record and escalation contacts with the handover, and set a review trigger for sustained growth or changing requirements.
