Jorge Ramon(@jrletosa) 's Twitter Profile Photo

Alex Monahan Robert Sahlin Mim Ideally you don't want your tasks to compete for the same resources airflow itself needs. It also allows for a more granular configuration of the resources. medium.com/bluecore-engin…

account_circle
Robert Sahlin(@robertsahlin) 's Twitter Profile Photo

Ananth Packkildurai Mim ADF is probably the worst service I’ve ever used in my 10 years as a data engineer. I celebrated when we replaced and sunset ADF in our pipelines.

account_circle
Robert Sahlin(@robertsahlin) 's Twitter Profile Photo

Rahul Jain Interesting, I find it awesome, but we stream data, even the data that is ingested via gcs. With BQ storage write api you can set up cdc as well. I have tested using CF and impressed by the performance.

account_circle
Turar(@abstractions360) 's Twitter Profile Photo

Robert Sahlin Ah I see thanks. So when you deserialize, the metadata piece is all the same format, and the payload is separate and can only be deserialized once you determine the message type?

account_circle
David Gasquez(@davidgasquez) 's Twitter Profile Photo

Robert Sahlin Ismael Ghalimi Mim Interesting! Never thought about this approach. Will try to test it out soon. I think it should be easy to have DuckDB queries running on Collab using BQ Storage API.

account_circle
David Gasquez(@davidgasquez) 's Twitter Profile Photo

Robert Sahlin Ismael Ghalimi Mim Couldn't resist trying out and was able to run a Collab with DuckDB querying BigQuery tables using the arrow interface!

If I'm correct, the Storage Read API is at least one order of magnitude cheaper than BQ, right?

account_circle