我应该对 I/O 和 CPU 绑定的任务使用多线程吗？

Question

There is a lot of information on the web about Python's multi-threading and multiprocessing modules and why someone should use one over the another. The basic knowledge is that multi-threading is ideal for I/O bound tasks and multiprocessing for CPU bound tasks. All good so far.

But real-life scenarios are often not that clear and often you are faced with situations when the underlying task is a mix of both I/O and CPU bound work. I'm currently facing a scenario where my task would be spending 80% of its time waiting for an API call that is essentially an I/O bound task but, it will spend 20% of the remaining time in processing the data from that API. This makes my task mixed with I/O and CPU bound work.

My question is that how should I decide the ideal module to be used for such scenarios? Should I also use a mixed model where the I/O bound work is done by multi-threading and vice versa?

Answer 1

There is a lot of information on the web about Python's multi-threading and multiprocessing modules and why someone should use one over the another. The basic knowledge is that multi-threading is ideal for I/O bound tasks and multiprocessing for CPU bound tasks. All good so far.

But real-life scenarios are often not that clear and often you are faced with situations when the underlying task is a mix of both I/O and CPU bound work. I'm currently facing a scenario where my task would be spending 80% of its time waiting for an API call that is essentially an I/O bound task but, it will spend 20% of the remaining time in processing the data from that API. This makes my task mixed with I/O and CPU bound work.

My question is that how should I decide the ideal module to be used for such scenarios? Should I also use a mixed model where the I/O bound work is done by multi-threading and vice versa?