Skip to content

8. Concurrency and Parallelism.md

Latest commit

 

History

History
75 lines (47 loc) · 3.38 KB

8. Concurrency and Parallelism.md

File metadata and controls

75 lines (47 loc) · 3.38 KB

Concurrency and Parallelism in System Design

  • Concurrency and Parallelism are two related but distinct concepts in system design and computer science.
  • Both deal with executing multiple tasks, but they approach it differently.

Concurrency

  • Concurrency refers to the ability of a system to handle multiple tasks at the same time conceptually.

  • Tasks make progress by being interleaved (not necessarily executed simultaneously).

  • Useful when tasks involve I/O-bound operations or when tasks must share resources.

Concurrency is achieved through the interleaving operation of processes on the central processing unit (CPU) or in other words by the context switching. It increases the amount of work finished at a time.

  • Context Switching is the process of storing and restoring the state (context) of a CPU such that multiple processes can share a single CPU resource. It allows multiple processes to share a single CPU, and is an essential feature of a multitasking operating system.

  • Key Points:

    • Focuses on dealing with lots of things at once.

    • Achieved using threads, async programming, or event loops.

    • Doesn’t always mean faster execution, but ensures responsiveness.

  • Example:

    • A chat application where multiple users send messages. Even on one CPU, the system interleaves tasks so all users feel responsive interaction.
    • A web server handling multiple client requests concurrently using threads or async calls.
    • Even if only one CPU core is available, concurrency is achieved through context switching.

Parallelism

  • Parallelism refers to actually executing multiple tasks at the same time on different processors/cores.

  • Involves true simultaneous execution of computations.

  • Best suited for CPU-bound tasks like mathematical computations, ML model training, big data analysis or simulations.

  • Key Points:

    • Focuses on doing lots of things at once.

    • Requires multiple CPU cores or distributed systems.

    • Improves throughput by splitting work across processors.

  • Example:

    • Splitting a large array into chunks and processing each chunk on a different CPU core simultaneously.
    • GPU-based deep learning where thousands of matrix operations run in parallel.
    • A machine learning training job that divides data across GPUs and processes them simultaneously for speed.

Concurrency vs Parallelism

Aspect Concurrency Parallelism
Definition Managing multiple tasks at once Executing multiple tasks at once
Execution Interleaved (not simultaneous) Truly simultaneous
Use Case I/O-bound tasks (web servers, DB ops) CPU-bound tasks (ML, simulations)
Hardware Need Can run on a single core (via scheduling) Needs multiple cores/CPUs/GPUs
Goal Responsiveness, resource sharing Speed and throughput

Why Does This Matter?

Understanding concurrency vs. parallelism helps you:

  • Choose the right design for performance and scalability.
  • Avoid common pitfalls like race conditions.
  • Leverage system resources effectively.