We will soon be living in an era in which we cannot guarantee survivability of any single point. However, we can still design systems in which system destruction requires the enemy to pay the price of destroying n of n stations. If n is made sufficiently large, it can be shown that highly survivable system structures can be built…
– Paul Baran, On Distributed Communications, Volume I, 1964.
Who invented packet switching? Like the development of hypertext, packet switching seems to have been an idea that wanted to be discovered. The packet switching concept was first invented by Paul Baran in the early 1960’s, and then independently a few years later by Donald Davies. Leonard Kleinrock conducted early research in the related field of digital message switching, and helped build the ARPANET, the world’s first packet switching network.
Baran invented the concept of packet switching while a young electrical engineer at RAND when he was asked to perform an investigation into survivable communications networks for the US Air Force, building on one of the first wide area computer networks created for the SAGE radar defence system. His results were first presented to the Air Force in the summer of 1961 as briefing B-265, then as paper P-2626, and then in 1964 as a series of eleven amazingly thorough, comprehensive papers titled On Distributed Communications.
Baran’s 1964 papers go well beyond documenting the breakthrough concept of packet switching and describe a detailed architecture for a large-scale, distributed, survivable communications network designed to withstand almost any degree of destruction to individual components without loss of end-to-end communications. Baran also assumed that any link of the network could fail at any time, and so the network was designed with no central control or administration.
Baran’s groundbreaking work helped to convince the US Military that wide area digital computer networks were a promising technology. Baran also talked to Bob Taylor and J.C.R. Licklider at the IPTO about the concept since they were also working to build a wide area communications network. Baran’s papers then influenced Roberts and Kleinrock to adopt the technology when they joined the the IPTO for development of the ARPANET, laying the groundwork that led to its incorporation into the TCP/IP network protocol used on the Internet today.
In one of many interesting such synchronicities in the history of science, Baran’s packet switching work was strikingly similar to the work performed independently a few years later by Donald Davies at the National Physical Laboratory, including common details like a packet size of 1024 bits. The term “packet switching” itself was taken from Davies work, since Baran had called the concept the bit less catching “distributed adaptive message block switching”.
At the time in the early 1960’s, existing communication networks were made from dedicated, analog circuits mainly used for voice telephone connections which were always on once activated. Packet switching completely changed this perspective by viewing networks as discontinuous, digital systems that transmit data in small packets only when required. At first glance this looks like it introduces two compromises in design:
- Discontinuity. It gives up the advantage of an always-on, continuous connection.
- Conversions. Analog communications like voice have to undergo analog-to-digital encoding to get onto the network and then digital-to-analog decoding at the destination to be read – extra work.
However, as always the details make the difference, and it turns out that packet switching introduces four practical advantages that far outweigh any hypothetical disadvantages:
- Digital. It makes communications digital, which means they can be made error free. It also means that communications from digital computers have no conversion overhead or transformation error.
- Processing. It moves the computer into the network by placing software systems at each node, which can then be upgraded and improved to enable the network to continually get better.
- Redundancy. It eliminates dependence on any one communication link, enabling the network to survive considerable damage.
- Efficiency. It enables more than one communication to share a given link at the same time, greatly increasing the number of total communications the network can support at any one time.
The established communications establishment — primarily telecommunications companies — was skeptical about the idea at first, but it was quickly shown that a packet switching network typically worked better, faster, and cheaper than a dedicated circuit network. Since the network shared all of the available bandwidth on a packetized basis, many communications could occur simultaneously. This was a major discovery, and the key concept that made wide-area communication networks and the Internet itself cost-effective and possible.
Interestingly, the development of packet switching came only a few decades after the development of quantum mechanics in physics, which began when Albert Einstein showed that waves of light could also be described as streams of individual photons. Despite this background, Baran says that his direct inspiration was Claude Shannon’s machine in which he trained a mechanical mouse to find its way through a maze as an existence proof. This led Baran to build on the idea and hypothesize that a message could be broken up into individual packets of information that could then find their own way to the destination through the network.
Resources. The following page provides a wealth of information about packet switching networks in the 1970’s: