Configure the compute size for a Databricks app
This feature is in Beta.
Each Databricks app runs on compute resources that determine its processing power and memory. Choose a compute size when you create or edit an app to match your workload requirements.
Available compute sizes
The following compute sizes are available for Databricks apps:
Size | CPU | Memory | Cost per hour | When to use |
---|---|---|---|---|
| Up to 2 vCPUs | 6 GB | 0.5 DBU | Standard apps with moderate resource needs, such as dashboards, simple data visualizations, and forms. Most apps work well with this size. |
| Up to 4 vCPUs | 12 GB | 1 DBU | Apps that process large datasets in memory, handle high concurrency, or perform more intensive computations. |
If you don't specify a compute size, Databricks assigns the Medium
size by default.
Configure compute size
Configure the compute size when you create or edit an app.
- Create app
- Edit app
To set the compute size when you create a new app:
- Click
Compute in the sidebar.
- Go to the Apps tab.
- Click Create app.
- In the Configure step, select a Compute size from the dropdown.
- Complete the remaining configuration steps and click Create app.
To change the compute size for an existing app:
- Click
Compute in the sidebar.
- Go to the Apps tab and click the app name.
- Click Edit
- In the Configure step, select a different Compute size from the dropdown.
- Click Save.
After you change the compute size, the app continues to run on the previous compute size until the update to the new compute size finishes. After the update finishes, the app switches over to the new size.
View the current compute size for an app on the Overview tab.
Best practices for compute sizing
When you choose a compute size for your app, consider the following best practices:
- Most apps work well with the default
Medium
size. Only chooseLarge
if you encounter performance issues or know your app has high resource requirements. - Test your app with the selected compute size in a development or staging workspace before you deploy it to production.
- Consider cost implications. The
Large
compute size costs more per hour.