How the internet works

Lesson 1 of 81964–19696 min

ARPANET and the packet

At 22:30 on 29 October 1969, Charley Kline, a student programmer in Leonard Kleinrock's lab at UCLA, started typing LOGIN on a terminal connected to a computer at the Stanford Research Institute. At the other end, programmer Bill Duvall watched for the characters to arrive.

The L arrived. The O arrived. Then the system crashed. The first message ever carried by the network that would grow into the internet was “lo”. About an hour later the full login went through.

The two letters mattered less than the way they travelled. They were not sent down a reserved line like a phone call. They crossed a network that cut data into small blocks and forwarded each one on its own.

The trouble with a phone line

The telephone network of the 1960s used circuit switching: before you could talk, the exchanges reserved a dedicated path from one end to the other, and it stayed yours until you hung up. That suits a voice call. It suits computers badly. Their traffic is bursty: a short flurry of data, then silence. A reserved circuit sits idle most of the time, and if one exchange on the path fails, the whole connection drops.

Two people, one idea

At the RAND Corporation, Paul Baran published On Distributed Communications in 1964, a series of memoranda describing a network with many interconnected nodes where message blocks could take any available route, so the system could keep working after heavy damage. In Britain, Donald Davies at the National Physical Laboratory arrived at a similar design independently in the mid-1960s and gave the small pieces the name that stuck: packets.

AR1R2R3R4R5B

Received: —

A message is split into packets. Each one finds its own way across the mesh, and the receiver puts them back in order.

Packet switching works in four moves. The sender splits a message into packets. Each packet carries a header with its source, its destination and enough information to put it back in sequence. Every node along the way stores and forwards: it receives a packet, decides where to send it next and passes it on. Packets of the same message may take different paths and arrive out of order; the receiver reassembles them. Because no one owns a line, thousands of conversations share the same links, each using capacity only when it has something to send. Engineers call that statistical multiplexing, and it is why the internet is cheap enough to exist. It also means no single failure takes everything down: if one link breaks, packets simply route around it.

Key terms

Circuit switching
A dedicated path is reserved end to end for the whole conversation, used or not.
Packet switching
Data is cut into packets that are forwarded independently over shared links.
Header
The control information at the front of a packet: source, destination, sequence and more.
Store and forward
Each node receives a whole packet, then decides where to send it next.
Statistical multiplexing
Many flows share one link, each taking capacity only when it has data to send.

Check what you learned

  1. 1.Why did packet switching suit computer traffic better than circuit switching?

  2. 2.What was the first message carried between UCLA and SRI on 29 October 1969?

  3. 3.Who gave the name “packet” to the small blocks of data?

Practise it for the exam

Sources