Frequently Asked Questions
General
What is Lakemeter?
Lakemeter is a sizing and cost-estimation tool for supported Databricks workloads. It turns usage assumptions into monthly and annual planning costs and exports the supporting calculations to Excel.
How should I interpret the cost estimates?
Lakemeter uses its loaded pricing data and the assumptions entered in each workload form. The results are planning-grade estimates, not billing guarantees. Actual costs may differ because of commercial terms and usage that does not match the modeled scenario. Verify important assumptions against the official Databricks documentation, the Databricks pricing page, and customer-specific terms.
Is Lakemeter an official Databricks product?
No. Lakemeter is an open-source Databricks Labs project built as a Databricks App. It is not an officially supported Databricks product.
Configuration
What do Cloud, Region, and Tier mean?
- Cloud identifies the selected deployment provider.
- Region identifies the pricing region used for rate lookup.
- Tier identifies the pricing tier used for SKU availability and rate lookup.
The options shown in Lakemeter determine what can be selected for an estimate. See Getting Started for the workflow and use the official Databricks documentation for current platform availability.
Which workload type should I choose?
Use the Workload Sizing Guides. The catalog maps each sizing need to its canonical Lakemeter guide, including Databricks Apps, AI Parse, and Shutterstock ImageAI.
What's the difference between Classic and Serverless?
For a Classic calculation, Lakemeter asks for instance and scale-out assumptions and models DBU and VM costs separately.
For a Serverless calculation, Lakemeter does not add a separate VM component. Compare the expanded calculations using the usage assumptions appropriate for each mode. See the Calculation Reference for the shared cost structure and the official Databricks documentation for current product behavior.
AI Assistant
What can the AI assistant do?
The assistant can create fully configured workloads from a natural language description, analyze your existing estimate for cost optimization opportunities, suggest complete multi-workload architectures for common patterns (like RAG chatbots), and answer general Databricks pricing questions. See the AI Assistant guide for conversation examples.
Can the AI assistant modify my existing workloads?
The assistant can propose new workloads and can analyze your existing ones, but it cannot directly edit workloads you've already created. To modify an existing workload, use the workload form in the UI.
Export & Pricing
What format does the export use?
Lakemeter exports to .xlsx (Excel) format. The file includes formula-based cells, color-coded headers, frozen panes, a workload summary, a Platform Add-on section, and a final estimate summary. You can open it in Excel, Google Sheets, or any spreadsheet application. See the Exporting guide for full details.
Can I apply my negotiated discount?
Yes. Configure global, category, or SKU-specific discounts in the estimate. The Excel export shows list and discounted DBU and DSU costs. Platform Add-ons use a separate negotiated add-on discount that is applied after the published uplift.
Why do some workloads show multiple rows in the export?
Some workloads contain separately priced components. AI Search can add reranker and storage rows. Lakebase can include compute, storage, PITR, and snapshot rows. Databricks Default Storage emits stored-data and operation rows. All emitted rows are included in the totals. See the Exporting guide.
How are DBU rates determined?
Lakemeter resolves the list rate for the selected SKU and estimate context from its loaded pricing data. Use the SKU Explorer to inspect available rates and the Calculation Reference to understand how a rate is applied. Verify current public pricing on the Databricks pricing page.
Troubleshooting
A workload type is grayed out — why?
Lakemeter only enables workload options compatible with the selected estimate context. If a workload is unavailable, review the cloud, region, and tier selections. Use the official Databricks documentation to confirm current product availability.
The cost seems too high or too low — what should I check?
Common things to verify:
- Hours/Month — 730 means 24/7 operation. For business-hours-only usage, ~176 hours (8 hrs × 22 days) is more realistic.
- Number of workers — Each worker multiplies both DBU and VM costs.
- Acceleration and mode options — Enable them only when they match the planned workload, then inspect the resulting DBU quantity.
- Serverless vs Classic — Serverless has no VM costs but higher DBU rates. Classic has both.
- Discount — Configure the intended discount in the estimate. The Excel discount cells remain editable for additional what-if analysis.