Which is better - multithreading or parallelism (Python)?

Question

Highload

Which approach to writing a project is more preferable - parallel or multi-threaded in the context of Python? input Output. What exactly is a task? I am writing a project in Python that uploads data, everything depends on several tens of thousands of requests in 5-6 hours. I lean more towards the parallel way than the multithreaded way (hello GIL). I need advice from an expert with experience in this. Have I correctly defined for myself the way to implement the task, or did I miss it all the same? Thanks in advance

Libraries:
Multiprocessing
Threading

Answer 1

asyncio has a convenient mechanism for running a task asynchronously in a separate process via run_in_executor() , so it's not such a hard choice.
Multiprocessing is faster, but there are problems with passing complex objects. In fact, only primitives, dictionaries, and lists can be passed back and forth. Things like numpy arrays have to be converted.
So if that's not a problem, multiprocessing rules.