Hints explain the controls beside each choice.
Which instance should I choose?
Choose a bottleneck. CPU suggestions add physical cores and can reduce RAM per worker. Driver-memory suggestions keep workers unchanged. Click “Use” to apply a choice. Benchmark your workload before deciding.
Which instances and regions are covered?
The catalog covers selected fixed CPU instance types, sizes large to 8xlarge, in US East (N. Virginia) and US West (Oregon). The count and snapshot date appear above the table. Some Databricks-supported types, such as C7i, R7g, I3en and R7iz, and sizes above 8xlarge are not in this catalog. Missing prices stay unknown. Workspace, Databricks Runtime, policy and AZ availability still need checking.
Can I mix ARM (Graviton) and x86 instances in one cluster?
Not in one cluster. Databricks does not support mixing Graviton and non-Graviton instance types, because each needs a different Databricks Runtime. Driver and worker types can differ within one architecture, and flexible node fallbacks must keep the same architecture. Separate job clusters in one job can use different architectures, and you can still compare an x86 baseline with an ARM candidate. Compute configuration ↗
Does the compliance security profile change my cost?
Yes. A workspace with the compliance security profile pays the Enhanced Security and Compliance add-on, published as a percent of Databricks product spend at list price before discounts. The base is the Databricks charge, not EC2. This app adds it to each compared cluster. The baseline uses the same setting because both clusters are modeled in one workspace. Databricks bills the add-on on all Databricks usage in that workspace, including serverless and SQL. The profile requires the Enterprise plan and limits compute to the instance types on its AWS commercial list (updated see Sources). It is typically needed for HIPAA, PCI-DSS or FedRAMP Moderate workloads. It is intended to be permanent and applies to the whole workspace. Workspaces with only enhanced security monitoring or automatic cluster update pay the same add-on without the instance limits; turn the profile off here and add the percent yourself. Mission Critical workspaces, AWS GovCloud, fleet types and flexible node fallbacks are not modeled. Compliance security profile ↗ · Platform add-ons pricing ↗
What do EC2, DBU and DB mean?
EC2 is the AWS machine charge. A DBU is a Databricks billing unit. DB is the Databricks charge: DBUs/hour × price per DBU. The compute total adds EC2 and DB charges.
Why can a Photon price be missing?
This snapshot has no direct quote for that instance, plan, compute type and engine. That does not prove Photon is unsupported. Review the proposed priced replacement before applying it.
Does the Databricks Runtime version change the estimate?
No. The public DBU rates used here are published per instance type, plan, compute type and Photon setting, not per Databricks Runtime version, so the hourly estimate is the same for any version. The version can still change how long a job takes, which is why break-even uses your measured job duration. In this app, job duration means how long a run takes; it never means the Databricks Runtime version.
How is Spot modeled?
Spot prices change and are not published as a fixed list, so you enter an assumed Spot price as a percent of On-Demand for workers. The driver stays On-Demand. The Databricks charge is the same for Spot and On-Demand. The same-type On-Demand reference shows the cost if Spot capacity is not available.
What does the workspace import do?
Run databricks clusters list-node-types --output json for your workspace and choose the file. The app reads it in your browser and limits choices to types your workspace lists for the selected region. Nothing is uploaded. The filter is not saved in share links. It does not check compute policies or live capacity.
How do I share a comparison?
Copy the address bar. It holds the region, plan, compute, Photon, compliance security profile, nodes, counts, filters, baseline and job duration, and the Spot percent only when you choose to share it. It does not hold an imported workspace file. The recipient sees prices from the snapshot their browser loads, which can be newer than yours. Use Export scenario for a fixed record.
How do I start over?
Click Reset all beside the pricing controls. It reloads the calculator with the opening defaults and clears filters, comparisons, baseline inputs, hints and the imported workspace file. Copy the address bar first if you want to keep the current comparison. The TokensAndTea logo opens the main website. The YouTube link in the header opens the channel.
What do the exports contain?
Export CSV saves the visible table rows with hardware, rates and hourly costs. Export scenario saves a JSON record (schema version 2) of your settings, nodes, exact hourly costs, baseline and sources. Neither is a cluster configuration. When a baseline is complete, it also records the baseline hourly total and the break-even result.
Can I check AZ capacity or create a cluster?
This app compares fixed classic-compute shapes. It does not check live AZ capacity or create clusters. Check your workspace, Databricks Runtime version and policy before launching. Export scenario saves a comparison record.
Tokens