On Distributed Communications Library
Paul Baran graduated from Drexel University in 1949 (then called Drexel Institute of Technology), with a degree in electrical engineering. Paul worked for Hughes Aircraft Company's systems group before joining RAND in 1959. While working at RAND, Paul matured his early ideas on computer failures and redundancy and conceived of the Internet and digital packet switching, the Internet's underlying data communications technology. His concepts are still employed today; just the terms are different. His seminal work first appeared in a series of RAND studies published between 1960 and 1962 and then finally in the tome "On Distributed Communications," published in 1964.
Upon graduating from Drexel Paul joined the Eckert-Mauchly Computer Company in 1949, where he did technical work on UNIVAC computers, the first commercial computers in the USA. In 1955 he married Evelyn Murphy, moved to Los Angeles, and worked for Hughes Aircraft on radar data processing systems. He obtained his master's degree in engineering from UCLA in 1959, while taking night classes. His thesis was on character recognition.
Pauls early work was associated with addressing failures in computers. Computers at the time were very unreliable because of vacuum tubes, large numbers of connections, and other fault sources. While working on these problems Paul became interested in the human brain and how neurons are interconnected in the brain. It was known that the brain was very resilient with enormous recovery abilities. These ideas stayed with him for a lifetime and led to significant insights and break throughs in communications and computer systems.
Paul joined RAND Corporation in 1959, and was tasked with designing a "survivable" communications system that could maintain communication between end points in the face of damage from nuclear weapons. At the time of the Cold War, most American military communications used high frequency connections which would fail for many hours / days / weeks / months / years by a nuclear attack. Paul decided to automate a previous RAND (Franklin R. Collbohm's) system with emergency communication over conventional AM radio networks and showed that a distributed relay node architecture could be survivable. The Rome Air Development Center soon showed that the idea was practical.
Using mini-computer technology, Baran and his team developed a simulation suite to test basic connectivity of an array of nodes with varying degrees of linking. That is, a network of n-ary degree of connectivity would have n links per node. The simulation randomly 'failed' nodes and subsequently tested the percentage of nodes that remained connected. The result of the simulation revealed that networks where n = 3 had a significant increase in resilience against even as much as 50% node loss. Baran's insight gained from the simulation was that redundancy was the key. His first work was published as a RAND report in 1960, with more papers generalizing the techniques in the next two years.
Comment: As part of AAS, the Hughes reliability team showed that six 9's five availability could be achieved with 3-4 computer nodes in a fault tolerant configuration. Other mission critical communities ignore the reliability calculations and instead clearly state that 3 failures or 4 failures must occur before an undesired event is triggered. These ideas are fundamental to mission critical / safety driven systems and are even found in our automobiles where the brakes use 2 redundant hydraulic cylinder nodes and then an orthogonal third node via the parking brake which usually uses a cable. So Paul was just being a great engineer solving a problem using engineering techniques in a healthy and robust environment.
After proving survivability Paul and his team needed to show proof of concept for this design so hat it could be built. This involved high level schematics detailing the operation, construction, and cost of all the components required to construct a network that leveraged this new insight of redundant links. The result of this was one of the first store-and-forward data layer switching protocols, a link-state/distance vector routing protocol, and an unproved connection-oriented transport protocol. Details of these designs can be found in the complete series of reports On Distributed Communications, published by RAND in 1964.
The design was contrary to current telephony design, placing inexpensive and unreliable nodes at the center of the network, and more intelligent terminating 'multiplexer' devices at the endpoints. In Baran's words, unlike the telephone company's equipment, his design didn't require expensive "gold plated" components to be reliable. The Distributed Network that Baran introduced was intended to route around damage. It provided connection to others through many points, not one centralized connection. Fundamental to this scheme was the division of the information into "blocks" before sending them out across the network. This enabled the data to travel faster and communications lines to be used more efficiently. Each block was sent separately, traveling different paths and rejoining into a whole when they were received at their destination.
Comment: Connections in large systems is a source of many failures. To minimize intermittent and hard connection failures the connectors were typically gold plated. The "gold plated" components to be reliable were not needed in this alternative system architecture.
After the publication of On Distributed Communications, Paul presented the findings of his team to a number of audiences, including AT&T engineers (not to be confused with Bell Labs engineers, who at the time provided Paul with the specifications for the first generation of T1 circuits which he used as the links in his network design proposal). The AT&T engineers scoffed at his idea of non-dedicated physical circuits for voice communications, at times claiming that Paul simply did not understand how voice telecommunication worked.
Comments: This is not unlike the arrogance of IBM and the FAA on AAS.
Donald Davies at the National Physical Laboratory in the United Kingdom also thought of the same idea and implemented a trial network.
Comment: The common thread is probably Hughes Aircraft which was developing the Air Defense Systems for NATO and the USA. Hughes Aircraft, RAND, and SDC worked very closely together on Air Defense Systems. All three were non profit with RAND being an FFRDC.
While Baran used the term "message blocks" for his units of communication, Davies used the term "packets" as it was capable of being translated into languages other than English without compromise. He applied the concept to a general-purpose computer network. Davies' key insight came in the realization that computer network traffic was inherently "bursty" with periods of silence, compared with relatively constant telephone traffic. It was in fact Davies' work on packet switching, and not Baran's, that initially caught the attention of the developers of ARPANET, at a conference in Gatlinburg, Tennessee, in October 1967. Baran was happy to acknowledge that Davies had come up with the same idea as him independently. In an e-mail to Davies he wrote
You and I share a common view of what packet switching is all about, since you and I independently came up with the same ingredients.
Leonard Kleinrock, a contemporary working on analyzing data flow, also reached similar conclusions when he developed a theoretical basis for the operation of packet networks in his proposal for a Ph.D. thesis in 1961. He published his ideas in that year. However, Kleinrock's contribution to packet switching is disputed by some, including Robert Taylor, Baran and Davies. The U.S. National Inventors Hall of Fame, which recognizes inventors who hold a U.S. patent of highly significant technology, records Paul Baran and Donald Davies as the inventors of digital packet switching.
In 1969, when the US Advanced Research Projects Agency (ARPA) started developing the idea of an inter-networked set of terminals to share computing resources, the reference materials which they considered included Baran and the RAND Corporation's "On Distributed Communications" volumes. The resiliency of a packet-switched network that uses link-state routing protocols used on the Internet stems in some part from the research to develop a network that could survive a nuclear attack.
Paul died in Palo Alto, California at the age of 84 on March 26, 2011. Upon his death James Thomson, the president of RAND, stated that "Our world is a better place for the technologies Paul Baran invented and developed, and also because of his consistent concern with appropriate public policies for their use." One of the fathers of the internet, Vinton Cerf, stated that "Paul wasn't afraid to go in directions counter to what everyone else thought was the right or only thing to do." According to Paul Saffo, Paul Baran also believed that innovation was a "team process" and he didn't seek credit for himself. On hearing news of his death, Robert Kahn, co-inventor of the Internet, said: "Paul was one of the finest gentlemen I ever met and creative to the very end."
Comments: This is a key statement: "Paul Baran also believed that innovation was a "team process" and he didn't seek credit for himself." Many organizations have different value systems and philosophies. The concept of "team process" and to not seek self credit was a key driving value and philosophy at Hughes. It was based on the simple idea that some problems are so complex it is impossible for a single human mind to to conceive, understand, or solve. The trick is to build an effective geshtalt and the key to that geshtalt is efficient and effective communications between the nones or minds in this case. It is amazing how things tend to naturally unfold.
The following documents are in the On Distributed Communications Library. They are offered because it is a nice example of going from concept to design. Few people know how to do this and only a handful of organizations know how to this - go from a clean sheet of paper to a design concept.
I. Introduction to Distributed Communications Networks
Paul Baran, RM-3420-PR
Introduces the system concept and outlines the requirements for and design considerations of the distributed digital data communications network. Considers especially the use of redundancy as a means of withstanding heavy enemy attacks. A general understanding of the proposal may be obtained by reading this volume and Vol. XI.
II. Digital Simulation of Hot-Potato Routing in a Broadband Distributed Communications Network
Sharla P. Boehm and Paul Baran, RM-3103-PR
Describes a computer simulation of the message routing scheme proposed. The basic routing doctrine permitted a network to suffer a large number of breaks, then reconstitute itself by rapidly relearning to make best use of the surviving links.
III. Determination of Path-Lengths in a Distributed Network
J. W. Smith, RM-3578-PR
Continues model simulation reported in Vol. II. The program was rewritten in a more powerful computer language allowing examination of larger networks. Modification of the routing doctrine by intermittently reducing the input data rate of local traffic reduced to a low level the number of message blocks taking excessively long paths. The level was so low that a deterministic equation was required in lieu of Monte Carlo to examine the now rare event of a long message block path. The results of both the simulation and the equation agreed in the area of overlapping validity.
IV. Priority, Precedence, and Overload
Paul Baran, RM-3638-PR
The creation of dynamic or flexible priority and precedence structures within a communication system handling a mixture of traffic with different data rate, urgency, and importance levels is discussed. The goal chosen is optimum utilization of the communications resource within a seriously degraded and overloaded network.
V. History, Alternative Approaches, and Comparisons
Paul Baran, RM-3097-PR
A background paper acknowledging the efforts of people in many fields working toward the development of large communications systems where system reliability and survivability are mandatory. A consideration of terminology is designed to acquaint the reader with the diverse, sometimes conflicting, definitions used. The evolution of the distributed network is traced, and a number of earlier hardware proposals are outlined.
VI. Mini-Cost Microwave
Paul Baran, RM-3762-PR
The technical feasibility of constructing an extremely low-cost, all-digital, X- or Ku -band microwave relay system, operating at a multi-megabit per second data rate, is examined. The use of newly developed varactor multipliers permits the design of a miniature, all-solid-state microwave repeater powered by a thermoelectric converter burning L-P fuel.
VII. Tentative Engineering Specifications and Preliminary Design for a High-Data-Rate Distributed Network Switching Node
Paul Baran, RM-3763-PR
High-speed, or "hot-potato," store-and-forward message block relaying forms the heart of the proposed information transmission system. The Switching Nodes are the units in which the camplex processing takes place. The node is described in sufficient engineering detail to estimate the components required. Timing calculations, together with a projected implementation scheme, provide a strong toundation for the belief that the construction and use of the node is practical.
VIII. The Multiplexing Station
Paul Baran, RM-3764-PR
A description of the Multiplexing Stations which connect subscribers to the Switching Nodes. The presentation is in engineering detail, demonstrating how the network will simultaneously process traffic from up to 1024 separate users sending a mixture of start-stop teletypewriter, digital voice, and other synchronous signals at various rates.
IX. Security, Secrecy, and Tamper-Free Considerations
Paul Baran, RM-3765-PR
Considers the security aspects of a system of the type proposed, in which secrecy is of paramount importance. Describes the safeguards to be built into the network, and evaluates the premise that the existence of "spies" within the supposedly secure system must be anticipated. Security provisions are based on the belief that protection is best obtained by raising the "price" of espied information to a level which becomes excessive. The treatment of the subject is itself unclassified.
X. Cost Estimate
Paul Baran, RM-3766-PR
A detailed cost estimate for the entire proposed system, based on an arbitrary network configuration of 400 Switching Nodes, servicing 100,000 simultaneous users via 200 Multiplexing Stations. Assuming a usable life of ten years, all costs, including operating costs, are estimated at about $60,000,000 per year.
XI. Summary Overview
Paul Baran, RM-3767-PR
Summarizes the system proposal, highlighting the more important features. Considers the particular advantages of the distributed network, and conuents on disadvantages. An outline is given of the manner in which future research aimed at an actual implementation of the network might be conducted. Together with the introductory volume, it provides a general description of the entire system concept.