Concurrency is not Parallelism

Misc
#1

Listened to this talk by Rob Pike twice to ensure I got it.

This is my version of what he says:

  • Concurrency is being able to break down a problem into smaller chunks and handling them.
  • Concurrency does NOT mean being able to run these chunks simultaneously. They could or could not. It doesn’t care.
  • Parallelism is when you can run these chunks simultaneously (on a multi-core machine).
  • Concurrency enables parallelism. If you can break a problem down into smaller chunks and handle them, given more cores you could make them run simultaneously.
  • Parallelism by itself does not enable concurrency.

That’s really it in a nutshell.

2 Likes
#2

From what I have seen, we run the concurrent tasks in parallel. What would be an example of not running the chunks simultaneously?
Aim for this weekend if to read Chapter 8 and 9 of the Go Programming Language :slight_smile:

#3

Also, I think we should consider reading Tannenbaum distributed systems principles and paradigms.

#4

Running them on a single core machine. Then only one thing can run at a time. You can achieve the same effect on a multi-core machine by setting NumCpus to 1.

Doesn’t have such good ratings on Amazon. We should check if Jeff Dean recommended some books on distributed systems.