Multiplexing & De-multiplexing

Computer Network & Network Security System — Transport Layer, NEC licence examination syllabus (Nepal Engineering Council).

Multiplexing & De-multiplexing

How does your OS know which of fifty open connections a packet belongs to?

Multiplexing: many applications on your device send data, and the transport layer combines everything down onto the network. De-multiplexing: at the receiving end, the transport layer takes incoming segments and sorts them out to the correct application, using the (IP, port) combination as the sorting key.

This is exactly why you can have Chrome, Spotify, and a game all using the internet simultaneously on one connection to your router — the transport layer is quietly demultiplexing every incoming packet to the right app, dozens of times per second.

💡 One-liner: "What information does de-multiplexing use to route a segment to the right process?" → The destination port number (combined with the socket's full 4-tuple for TCP).

Two values for UDP, four for TCP

The chapter mentions the 4-tuple in passing, but the contrast is the substance of the topic and questions target it directly.

🔀 What each protocol sorts on

UDP — 2 valuesDestination IP and destination port. Two datagrams from entirely different senders reaching the same port go to the same socket.

TCP — 4 valuesSource IP, source port, destination IP, destination port. Every distinct connection gets its own socket, even when the destination port is shared.

This is what lets a single web server on port 443 hold thousands of simultaneous connections. Every client arrives at the same destination port, but each has a different source IP and source port, so all four values together are unique per connection — and the server can tell every client apart without needing a separate port for each.

💡 The exam question: "a server has one process listening on port 80, so how can it serve many clients at once?" The answer is the 4-tuple. The listening socket accepts connections; each accepted connection becomes its own socket identified by all four values, so segments never go to the wrong client.

Why UDP's simpler rule is a design choice

UDP is connectionless, so there is no connection to identify — a datagram is a standalone message. Sorting only by destination is exactly right for that model, and it is why one UDP socket can receive from many senders. The application, not the transport layer, works out who sent what.

That suits broadcast and multicast naturally, where a single socket should receive from many sources. Forcing TCP's 4-tuple rule onto UDP would break precisely that.

💡 Port numbers are 16 bits, so 0–65535, and the well-known range 0–1023 is reserved for standard services — 80 for HTTP, 443 for HTTPS, 22 for SSH. Outgoing connections take an ephemeral port from a high range instead, which is the source port that makes each client's 4-tuple unique.

Syllabus points

Multiplexing / de-multiplexing concept

Create a free account to tick topics off, take notes as you read, watch the video lessons and get a day-by-day study plan built around your exam date.