How does the handling of the data work in the long run ?

[Mo379]

Say the project is completed and the goal now is to begin training, how will the data be handled in that case ?

• will it be pre processed somewhere before and downloaded onto the training cluster for training ?
• will it be streamed bit by bit (streaming) ?
• what are the expected costs ?

Of course I dont expect that the questions are known but just as an estimate from anyone's previous experience.

also I'm not sure if my questions are naive so please feel free to lecture me here.