Skip to main content

Sizing Quick Reference

Use this page when reviewing an estimate. Workload-specific fields and formulas live in the Workload Sizing Guides, so they can change independently without making this reference stale.

Build an estimate

  1. Set the deployment context. Select the cloud, region, and pricing tier available for the scenario.
  2. Choose a workload. Use the workload catalog to match the billing pattern you need to size.
  3. Enter representative usage. Model expected runs, runtime, active hours, capacity, tokens, pages, images, or storage.
  4. Inspect the breakdown. Confirm the SKU, billing quantity, list rate, and separate infrastructure or storage components.
  5. Export and review. Treat the result as a planning estimate and validate important assumptions with the current Databricks pricing information and your commercial terms.

Common sizing quantities

QuantityHow to use it
Runs and runtimeEstimate how often a workload executes and how long each execution consumes compute.
Active hoursUse the expected billable operating time, not the total hours in a month unless the workload is modeled as always on.
CapacitySelect the compute, endpoint, warehouse, or database capacity appropriate for the expected load.
Scale-out countInclude workers, clusters, endpoints, nodes, or apps that can consume capacity concurrently.
Tokens or itemsEnter monthly input/output tokens, pages, images, or another workload-specific billing quantity.
StorageInclude persistent data and any separately modeled protection or overflow storage.

Key cost terms

TermMeaning in Lakemeter
DBUA Databricks billing unit. Lakemeter multiplies estimated DBU consumption by the selected SKU's list rate.
DSUA Databricks Storage Unit. Lakemeter multiplies DSU consumption by the exact regional DATABRICKS_STORAGE list rate.
SKUThe priced Databricks product entry used for a calculation.
List rateThe rate loaded for the selected SKU, cloud, region, and tier.
DiscountA planning adjustment applied to eligible list-rate costs. Confirm actual commercial terms separately.
VM costCloud infrastructure cost shown separately for calculations where Lakemeter models it independently.
Direct costA cost entered or calculated directly rather than derived from DBUs, DSUs, or separately modeled VMs.
Product Spend at ListDBU and DSU list cost before discounts; the basis for Platform Add-ons. VM cost is excluded.

Review checklist

  • Does the selected workload match the billing pattern being estimated?
  • Are usage values monthly and in the units shown by the form?
  • Does the scale-out count represent peak concurrency or average concurrency?
  • Are always-on and scale-to-zero assumptions intentional?
  • Are storage and data-protection quantities included where relevant?
  • Does the expanded calculation show the expected SKU and billing unit?
  • Have list rates and negotiated terms been checked before external use?

Go deeper