Cloud cost checklist: compute, storage, transfer and credits

Describe one workload and its owner

Begin with a concrete service such as a hypothetical customer portal with an application, database, uploaded files and scheduled reports. Record the expected users, operating hours, location requirements and recovery objectives. Then name the people responsible for deployment, access, updates, backups and support. A platform comparison becomes useful when both proposals describe the same workload. A free account offer or a large service catalogue does not define what that workload will require.

Map the proposal to named services and units

List the service, region, configuration, billing unit and quantity behind each row of an estimate. Compute might be described through instance time, a selected capacity or metered execution; a data service may charge for processing, reserved capacity or several separate components. Preserve the provider’s actual units rather than translating everything into an unsupported monthly seat price. Ask the person preparing the proposal to explain which resources are included and which depend on another account or supplier.

Include storage operations and copies

Record the working dataset, retained history, backups, replicas and temporary outputs. Identify which copies belong to the application and which are created by the proposed operating arrangement. Check storage class, request operations, retrieval, minimum retention and any required movement between services against the current service terms. A low capacity price cannot establish the cost of reading or restoring data. Define a representative recovery task and include the resources needed to complete it.

Draw the movement of data

Sketch the route from a user or source system into the application, through processing and storage, and back to its destination. Mark region boundaries, connected providers and exports. Request the relevant transfer and network-processing assumptions for each route. Do not assume that every movement inside a provider’s brand has the same charging conditions. Keep a second estimate for migration or a busy export period so that a one-off movement is visible rather than hidden inside a normal-month average.

Separate promotional access from retained cost

Record the eligibility, expiry and covered services of a trial or credit. Then model the same workload after the promotion. A recurring allowance can have service-specific limits; a sign-up credit can end while the workload remains. Write down whether the selected account can continue to incur charges, what controls are configured and who receives alerts. An alert and a hard spending stop are different controls. Verify the behaviour of the particular service before relying on either.

Measure the proposed arrangement before committing

Use an authorised small environment with an approved budget and synthetic data. Run the defined workload, a scheduled task and the planned cleanup procedure. Reconcile the observed service usage with the estimate, keeping the date, configuration and account scope. AWS Cost Explorer supports cost and usage investigation; Google Cloud supports billing exports for more detailed analysis. Reporting can involve its own storage or query costs. These tools support investigation; an estimate or short evaluation does not establish a guaranteed production bill.

Keep operating work and commitments visible

Add support, monitoring, backups, credentials, incident ownership and the internal time needed to maintain the environment. For an offered commitment or reservation, record what is being purchased, its term and the workload assumption behind it. Ask what happens when demand or architecture changes. Compare the full proposal with the team’s responsibilities rather than assuming a managed label transfers every task. Keep a named owner for releasing unused resources and reviewing the bill after the initial rollout.

Use the appropriate platform comparison

Start with AWS vs Azure, Google Cloud vs Oracle Cloud Infrastructure or DigitalOcean vs Akamai Cloud for the platform scope. For analytical workloads, compare Snowflake vs Databricks, BigQuery vs Redshift or Microsoft Fabric vs Teradata Vantage. Retain the workload brief and dated service estimate beside your comparison scorecard.

Frequently asked questions