Databricks Community

zach · ‎06-08-2022

Databricks uses DBU's as a costing unit whether based onto of AWS/Azure/GCP and I want to know if Databricks has a google cloud Big Query equivalent of --dry_run for estimating costs? https://cloud.google.com/bigquery/docs/estimate-costs

-werners- · ‎06-09-2022

Not that I know of.

Google uses number of bytes read to determine the cost.

Databricks uses DBU. The number of DBU's spent is not only dependent on the amount of bytes read (the more you read, the longer the program will run probably), but also the type of VM used.

Then there is also autoscaling which makes it harder to predict a price.

Also the total cost is not only DBU but also the provisioning cost of the VMs.

So that makes it pretty hard to predict a cost.

It would of course be very cool to have such a prediction.

zach · ‎06-09-2022

Hi @Werner Stinckens thank you for taking the time to reply and for the thoughtful response. I find it hard to believe that so many companies are using the type of compute when the price is hard to know. I understand there is some ambiguity with the bytes read and cluster type, do you know of a way to give a rough estimate?

-werners- · ‎06-09-2022

Databricks does give you a view on how many DBUs/hour a cluster consumes (from-to interval in case of autoscaling), see the cluster pane for this.

With that and a duration of the job, you can make an estimate. But the duration... for that you need to run the program (perhaps on small data and extrapolate).

This is a pretty rough estimate though. Maybe others have succeeded in doing this.

Kaniz_Fatma · ‎06-13-2022

Hi @zach welshman , We haven’t heard from you on the last response from @Werner Stinckens, and I was checking back to see if you have a resolution yet. If you have any solution, please share it with the community as it can be helpful to others. Otherwise, we will respond with more details and try to help.