Lecture 3 - Parallel Programming with Python

Question 1

Why does python have problem with multiprocessing?

Answer 1

Python uses GIL (Global Interpreter Lock) that ensures only one thread executed at a time

Question 2

What is multithreading?

Answer 2

• An operating system mechanism for parallel operations
• Threads run on the same CPU but utilize multiple cores

Question 3

What is multiprocessing?

Answer 3

Several independent processes (programs) are started by the main application process
- Subprocesses cannot share memory and usually don’t communicate with each other
- Can run on different CPUs

Question 4

Why is multiprocessing enough for data science?

Answer 4

• Many tasks are parallel
• There is too much data to process sequantially
• Dividing the data into subsets can reduce the time needed
• No communication needed between cluster nodes