COMP 150-IDS: Internet Scale Distributed Systems (Spring 2015) A Very Brief History of Early Digital Networking Noah Mendelsohn Tufts University Email: noah@cs.tufts.edu Web: http://www.cs.tufts.edu/~noah Copyright 2012 & 2015 – Noah Mendelsohn Shannon & Information Theory 2 © 2010 Noah Mendelsohn Claude Shannon and Information Theory 1948: Claude Shannon publishes: A mathematical theory of communication* * http://cm.bell-labs.com/cm/ms/what/shannonday/shannon1948.pdf 3 Photo by Tekniska Museet © 2010 Noah Mendelsohn Claude Shannon and Information Theory Shannon’s work is as fundamental to digital communication as Turing’s is to digital computing Information theory – Quantifies information: how much information does a bit represent? – Relates information transmission to bandwidth requirements – Provides quantitative analysis of rate at which information can be sent over a noisy channel Shannon showed that information could be communicated reliably He predicted how much information could be communicated reliably given that channel characteristics are known BTW: Shannon and Turing knew each other and met for several months 4 © 2010 Noah Mendelsohn Whirlwind, SAGE & US Air Defense 5 © 2010 Noah Mendelsohn Early history of digital data transmission 1948: Claude Shannon publishes: A mathematical theory of communication* Late 1940’s: US seeks means of providing cold-war air defense Late 1949: Digital Radar Relay – experiment sending radar dataover phone lines - first digital transmisison over the phone 1951: MIT Whirlwind machine goes online 6 © 2010 Noah Mendelsohn Whirlwind computer 7 © 2010 Noah Mendelsohn Whirlwind computer The first significant real time computer system Innovation: core memory & digital networking 5000 vacuum tubes 16 bit parallel ALU 20,000 instructions/second – limited by storage speed 4000 bytes of core memory – invented for Whirlwind Pictures by Dan Smity 8 © 2010 Noah Mendelsohn Core Memory Aside: for 20 years before transistor memories became available, core memory made digital computing practical 9 © 2010 Noah Mendelsohn Early history of digital data transmission 1948: Claude Shannon publishes: A mathematical theory of communication* Late 1940’s: US seeks means of providing cold-war air defense Late 1949: Digital Radar Relay – experiment sending radar dataover phone lines - first digital transmisison over the phone 1951: MIT Whirlwind machine goes online (approximate) 1953: Cape Cod System tests sending radar data through phone lines to Whirlwind 10 © 2010 Noah Mendelsohn Forrester promptly began preparing to receive and process digitized radar signals. The feasibility demonstration of the radar/digital-data concept took place at Hanscom Field in September 1950. The radar, which was an original experimental model of a microwave earlywarning unit built by the wartime MIT Radiation Laboratory, closely resembled the radars used in the D-Day invasion of Normandy. History of Whirlwind and MIT Lincoln Lab: http://www.ll.mit.edu/about/History/origins.html 11 © 2010 Noah Mendelsohn Forrester promptly began preparing to receive and process digitized radar signals. The feasibility demonstration of the radar/digital-data concept took place at Hanscom Field in September 1950. The radar, which was an original experimental model of a microwave earlywarning unit built by the wartime MIT Radiation Laboratory, closely resembled the radars used in the D-Day invasion of Normandy. While military observers watched closely, an aircraft flew past the radar, the digital radar relay transmitted the signal from the radar to Whirlwind via a telephone line, and the result appeared on the computer's monitor. The demonstration was a complete success and proved the feasibility of ADSEC's air defense concept. History of Whirlwind and MIT Lincoln Lab: http://www.ll.mit.edu/about/History/origins.html 12 © 2010 Noah Mendelsohn Early history of digital data transmission 1948: Claude Shannon publishes: A mathematical theory of communication* Late 1940’s: US seeks means of providing cold-war air defense Late 1949: Digital Radar Relay – experiment sending radar dataover phone lines - first digital transmisison over the phone 1951: MIT Whirlwind machine goes online (approximate) 1953: Cape Cod System tests sending radar data through phone lines to Whirlwind 1957: First SAGE system, based on Whirlwind technology – SAGE runs US air defenses until 1983! (video) * * Good book on Whirlwind: Bright Boys, by Tom Green History of Whirlwind and MIT Lincoln Lab: http://www.ll.mit.edu/about/History/origins.html 13 © 2010 Noah Mendelsohn The SAGE Computer System (AN/FSQ-7) 60,000 vacuum tubes[8] (49,000 in the computers) Consumed up to 3 megawatts of electricity Performed about 75,000 instructions per second Memory: ~65,000 32 bit words Source: http://en.wikipedia.org/wiki/AN/FSQ-7_Combat_Direction_Central 14 © 2010 Noah Mendelsohn Paul Barran & Packet Switching 15 © 2010 Noah Mendelsohn Paul Baran, Donald Davies and Packet Switching 1964: Paul Baran proposes packet switching design Design goal: a resilient network to maintain command and control 16 © 2010 Noah Mendelsohn Circuit switching (the way the old phone system worked) When you make a call… …switches are set to reserve links for a fixed route for the life of the call 17 © 2010 Noah Mendelsohn Packet switching When you communicate… …packets find independent routes through the network 18 © 2010 Noah Mendelsohn Packet switching When you communicate… …packets find independent routes through the network 19 © 2010 Noah Mendelsohn Paul Baran, Donald Davies and Packet Switching 1964: Paul Baran proposes packet switching design Design goal: a resilient network to maintain command and control Questions to consider: – Performance: better or worse than circuit switch? – How are routing tables maintained? – Why was it counter-intuitive Success of early packet switching tests motivates government funding for ARPANet 20 © 2010 Noah Mendelsohn Packet vs. Circuit Switching Circuit switching – Good for continuous predictable flows – Easy to put “smarts” into the middle of the network (smart switches) – The way to go when all you have is analog communication Packet switching – Adapts well to changing loads – Relatively cheap to make lots of quick “connections” – Paul Baran’s insight: digital makes packet switching possible (packet does not “degrade” as it gets copied through intermediate nodes) 1960’s: AT&T did not believe packet switching would work Packet switching tends to put value outside the network When systems are fault-tolerant, you can often build them from cheaper components 21 © 2010 Noah Mendelsohn A Brief History of The Internet 22 © 2010 Noah Mendelsohn History of the Internet DNS (Mockapetris) Baran & Davies Packet Switching Work Prompts Gov’t Investment 14 Arpanet Nodes + 1/month NSFNet 1989: 80,000 hosts Tim BL Proposes Web ARPANet Developed 1960 Bob Metcalfe Begins Ethernet work at Xerox PARC First Web Server 1965 1970 TCP/IP (Cerf & Kahn) 1975 1980 1985 1990 Adapted from http://www.computerhistory.org/internet_history/ 23 © 2010 Noah Mendelsohn History of the Internet DNS (Mockapetris) Baran & Davies Packet Switching Work Prompts Gov’t Investment 14 Arpanet Nodes + 1/month 24 NSFNet Metcalfe Begins Ethernet work at By 1992, the Internet is doubling Xerox PARC 1989: 80,000 hosts Tim BL Proposes Web in size every 3 months ARPANet Developed 1960 First Web Server 1965 1970 TCP/IP (Cerf & Kahn) 1975 1980 1985 1990 © 2010 Noah Mendelsohn