Understanding Network I/O: Blocking, Non‑Blocking, Async & Multiplexing Explained

It is often said that reviewing the past helps us learn anew; in work we frequently hear colleagues say, “I know this, but I forgot the details, I need to Google it.” Knowledge should form a mental system rather than relying on constant searches. This article provides a concise summary of basic network I/O concepts—synchronous vs asynchronous and blocking vs non‑blocking—to refresh and deepen understanding.

Preface

In network programming, terms such as blocking, non‑blocking, synchronous, and asynchronous are frequently mentioned. What exactly are synchronous I/O, asynchronous I/O, blocking I/O, and non‑blocking I/O, and how do they differ?

Data Flow

For a network I/O operation (using read as an example), two system entities are involved: the process (or thread) that initiates the I/O and the kernel. When a read occurs, it goes through two stages:

Waiting for the data to be ready.

Copying the data from the kernel to the process.

Remembering these stages is important because the differences among I/O models appear in how they handle each stage.

Network I/O Model Detailed Analysis

Common I/O models include blocking, non‑blocking, I/O multiplexing, and asynchronous I/O.

1. Blocking

Imagine going out to eat on a busy weekend. After ordering, you must wait in the restaurant until the food is ready before you can continue. This waiting period represents blocking I/O.

2. Non‑Blocking

Instead of waiting, you repeatedly check with the server whether the food is ready, going back and forth many times. This polling behavior illustrates non‑blocking I/O, where the process must keep checking the I/O status.

3. Multiplexing

With an electronic display showing order status, you can glance at the screen instead of repeatedly asking the server. This is analogous to I/O multiplexing (e.g., select , poll , epoll ).

4. Asynchronous

You place a delivery order by phone and then relax at home. When the food arrives, you are notified without having to check. This exemplifies asynchronous I/O, where the kernel handles the operation and notifies the process upon completion. Linux provides AIO library functions for async I/O, though they are rarely used; popular async libraries include libevent, libev, and libuv.

Synchronous vs Asynchronous

Synchronous I/O means the process initiates the operation and then waits (or polls) until it completes. Asynchronous I/O means the process initiates the operation and immediately returns to do other work; the kernel completes the I/O and later notifies the process.

Blocking vs Non‑Blocking

Simply put, if an operation cannot return immediately and forces the process to wait, it is blocking; if the operation returns immediately and the process can continue, it is non‑blocking.

Summary

From the diagrams it is clear that non‑blocking I/O and asynchronous I/O differ significantly:

In non‑blocking I/O, the process rarely blocks but must actively check the I/O status and, once data is ready, must explicitly call recvfrom to copy the data to user memory. Asynchronous I/O, on the other hand, hands the entire operation to the kernel; the kernel completes it and notifies the process, eliminating the need for the process to poll or copy data manually.