Hello, @FanMichelle0729!

It depends on the type of Compute you're using in Databricks. Non-serverless compute runs in your external cloud account (e.g., AWS) and becomes unusable if that account is deactivated.

On the other hand, serverless Compute runs in a Databricks-managed environment and doesn't rely on an external cloud account (AWS, Google Cloud, Azure). This means it remains usable even if your external cloud account is deactivated.

From Databricks architecture , the serveless compute means SQL Datawarehouse serveless ? and the classical compute means the all-purpose compute (ex. picture 2) ?

https://docs.databricks.com/en/_images/architecture.png 

picture2

@FanMichelle0729 

Databricks has implemented serverless capabilities not only for SQL Warehouses but also for All-Purpose Clusters, Job Clusters, and DLT.

However, in the image you attached, the All Purpose Compute tab only displays non-serverless compute options, while the SQL Warehouse tab shows both serverless and non-serverless options.

@Takuya-Omi 

Thanks for replying , about the image i attached , It's my first to try Databricks and connect AWS account. 

So it means when default use Databricks it will provide non-serverless (classical serveless) compute , right ?

@FanMichelleTW 

No, Databricks recommends using serverless compute, and you can use serverless compute as well.

To do so, open a notebook and check the top-right corner to see if a serverless compute option is in a Ready state. If it is, simply select it, and you can start using it immediately.

The "All-Purpose Compute" tab in the compute section does not display serverless compute, but rest assured that it runs on Databricks-managed infrastructure in AWS.