Chapter 19 ---> Thread Pooling

Threads can impact performance when:
- linear increasing of context switch cost, threads creation/destruction cost.
- nonlinear increasing of cache missing.
To mitigate problem, BCL makes Thread Pool. It can:
- reuse threads for different tasks (this is also why task and thread become separate concepts)
Pitfalls:
- Thread Pool requires tasks to finish in seconds, otherwise the benefit will be gone.
- Can't refer to a task given to it. So can't sync with tasks. That's why we have to use Sleep() to sync.
CPU bound short-term task -> thread pool
I/O bound or long-term task -> TPL