What is most reasonable way to read data out of Google Cloud Storage?

[feiyao]

Hi community,

I have recently gone through TensorFlow example workflows (https://developers.google.com/earth-engine/guides/tf_examples). By doing a wee bit more authentication, I have successfully adapted and run these workflows on my local machines. However, I am unclear whether using tensorflow i/o tools and/or other i/o tools (e.g. rasterio) to read data from Google Cloud Storage directly to local machines' memories counts as downloading it and thereby will incur expensive Google Cloud Storage Network fee (https://cloud.google.com/storage/pricing#network-pricing). My concern largely comes from the fact that I have found a fee called "Download Worldwide Destinations (excluding Asia & Australia)" in my month bills for those days I adapted and run the above TensorFlow example workflows.

If reading and/or downloading data from Google Cloud Storage directly to local machines's memories is NOT free of charge, then what is the most reasonable way to achieve the same effects? What's come into my mind is via Google Drive plus Compute Engine or Colab: 1) Is copying data from Google Cloud Storage to Compute Engine in the same locations free of charge? Is downloading data from Compute Engine to Google Drive free of charge? Is downloading data from Google Drive to local machines free of charge? 2) Is copying data from Google Cloud Storage to Google Drive via Colab free of charge? Is downloading data from Google Drive to local machines free of charge?

Any insights would be highly appreciated. Thanks in advance!

All the best,
Fei

[comaro]

As seen here, when data is moved from different locations in the same continent, it is not considered to be part of the 'free' package.