A BRIEFING ON THE DISTRIBUTED ADAPTIVE MESSAGE-BLOCK NETWORK Paul Baran April 1965 P-3127 A BRIEFING ON THE DISTRIBUTED ADAPTIVE MESSAGE-BLOCK NETWORK Paul Baran The RAND Corporation Santa Monica California LIMITATIONS OF CURRENT COMMUNICATIONS We are witnessing evolution of our national military doctrine away from massive reSponse -where only the go word need be transmitted- toward a more flexible reSponse This change creates new problems for the command and control system designer Figure sketches the increased command and control in communications implied by flexible response doctrines As we move to the right of the scale increasing flexibility of response the amount of military communications that HERE survive attack markedly increases The degree of flefibility of response possible may be limited primarily by the avail- ability of survivable communications If this be true our choice of military doctrine may be dictated more by the availability of communications than by desire We feel that communications available to the military may fall short in several resPects see Fig 2 These include Any views expressed in this paper are those of the author They should not be interpreted as reflecting the views of The RAND Corporation or the official opinion or policy of any of its governmental or private research sponsors Papers are reproduced by The RAND Corporation as a courtesy to members of its staff This paper was presented to The RAND Corporation's Air Force Advisory Group and Board of Trustees meeting November 1964 DEMAND FOR COMMUNICATIONS MASSIVE ALTERNATIVE DYNAMIC CONTROLLED RESPONSE WAR PLAN RETARGETING ESCALATION GO WORD - - FIXED RESPONSE FLEXIBLE RESPONSE Fig 1 Command communications requirements posed by a flexible response doctrine SUQVIVABILITY secoecv 0 20000 DELAY TIMES 0 VOLUME 0 FLEXIBILITY useo-To-useo 6 COST Fig 2 How is command and control for flexible response limited by communications Survivability Existing communications networks are for the most part highly vulnerable to overt and covert attacks directed against communications Secrecy Most of our military communication is trans mitted with negligible protection to the loss of critical information by an eavesdropper Only a small portion of military communications is protected and the military is often highly constrained in subjects discussed over the telephone for fear of eavesdropping Error Rate Quality Those who must connect computerS with other computers or with remote entry devices are not completely satisfied with the high error rate of the present day communication networks Without great care and much ancillary equipment computers are very intolerant of errors caused by the communications links Delay Times Partially because of the high costs of communication military hard copy paper traffic is primarily a center-to center operation with long delay times between end users Flexibility User to User We lack the ability to achieve complete flexibility of connection between users particularly where secrecy is mandatory We would prefer to be able to speak to anyone quickly without making prior arrangements because we often do not know in advance to whom we wish to Speak under a completely flexible response doctrine--almost by definition Volume We note a rapidly increasing volume of military communications in the future Our present communication plant is highly limited relative to future demands in the volume of traffic that it is able to transmit We feel that the military would be more effective if communications were not always treated as a scarce resource Cost Because of our methods of assigning cost it is difficult to determine exactly what we are paying for military communications One recent estimate held that we were spending about a billion dollars per year and the cost increases about 15 per cent anually Future communications costs will not be cheap BEATING THE VULNERABILITY PROBLEM Figure 3A shows the centralized network -one method of building communication networks used in the past Here a group of stations nodes wishing to inter-communicate with one another are all tied to a central switching node This central switching node establishes paths of communication from any node to any other node Such networks require only extremely simple equipment but suffer from the disadvantage of being highly vulnerable Destruction of a single central switching node destroys communications for the_entire network With the decentralized network in Fig 3B we reduce the severity of this problem by connecting each of the original group of nodes to a few nearby stations which act as sub-switching centers These sub-switching centers are connected Destruction of a single node no longer destroys all communication However the network is still less than wholly survivable since the destruction of a few of the sub switching centers destroys network communications In this briefing we consider in detail the use of the distributed network -a network in which each station is connected gnly to its nearest neighbors Fig 3C We show that such a network is potentially more survivable than that in the previous example We then consider the practical problem of how to build switching gadgets that allow any station to talk to any other station by passing through a large number of nodes in tandem SURVIVABILITY OF THE DISTRIBUTED NETWORK CONFIGURATION A Monte Carlo computer simulation is used to de termine how the distributed network fares under attack A wa q01d muqueulnA au uneag g 0 8 MUOMHN MOOMLBN group of nodes as in Fig 4A is subjected to attack in which a certain percentage of the nodes is destroyed on a random basis The number of nodes still able to communicate with one another is computed for two variables the probability of destruction of a single node and the degree of connectivity of the network All measurements shown are for an 18x18 array of 324 nodes Figure AB shows the definition of the degree of redundancy or the connectivity used If for example each node connects to only two other stations it is defined as a network of redundancy level one If twice as many links are used as this minimum possible connectivity the network is said to have re- dundancy level two Examples are shown for redundancy levels of 1 1-1 Throughout the exercise it is assumed that a perfect switching ability exists This allows all nodes to maintain communication with others regardless of the number of tandem nodes traversed Figure 4C demonstrates the high payoff for redundancy The vertical axis represents the fraction of nodes that have withstood the physical attack and are in electrical communications with one another The horizontal axis is the specified single node probability for destruction of a raid directed against the network Figure 40 shows that low-redundancy networks fall apart under light attack However if the redundancy level is increased to values of only three or four extremely tough networks can be built By tough we mean that if a station survives the direct attack it has a very good chance of being in communication TRACTION OF STATIONS IN COMMUNICATION BEST POSSIBLE LINE 1 0 SINGLE NOTE PROBABILITY 0F OESTRUCT ION Fig 4 Survivability of a distributed network with all the other surviving stations of the network Two key points are to be noted from Fig 40 First only a 'moderate amount of redundancy is required--on the order of three or four More redundancy buys little Secondt such networks are able to maintain good post-attack communications even where perhaps half of the stations are destroyed Thus it appears that the distributed network configura- tion is a good one to consider when we must build networks able to survive heavy attack PROBLEMS OF BUILDING A DISTRIBUTED NETWORK To take advantage of the high survivability theo- retically possible in distributed networks we have to cross many tandem switching centers This is difficult in present-day analog-type networks because quality deteriorates in the tandem connections We compound this problem if we wish to use a wide Spectrum of emergency communication links while still trying to provide a high- quality high-reliability system Our belief that we should be able to use unreliable links stems from the viewpoint that it should not make much difference whether unreliability comes from enemy attack or from local electronic causes In a future' emergency we might even wish to include television broad- casting links as well as conventional microwave links in our communications network A low-altitude satellite is an example of an inherently unreliable link When the satellite is overhead we have a link when the satellite is over the horizon the link is out From the system's point of view this is just another unreliable link In -10- the distributed network the effective overall increase of system reliability depends upon the extensive use of unreliable elements To meet the reliability problem caused by many un- reliable links in tandem we must do two things 1 use digital modulation and 2 employ a store-and forward information transfer The only way that we know how to transmit signals via a large number of tandem repeaters without having the signal irrevocably corrupted is to use digital modulation As long as the signal is even stronger than the noise we may re form the signal before its next transmission hop Recently developed digital computer techniques will also allow powerful processing of the digital stream for error detection secrecy encoding and automated switching Thus the network of the future that we will describe will use all digital transmission It will however as will be shown handle conventional analog igpgt signals as well We call it an all-digital system because all the proces- sing and transmission in the network is conducted in an all-digital manner In the network to be described it may be necessary to use links of different data rates being fed from a variety of different users operating with widely different Adata rates Therefore we will design our entire system around a standard package of bits which we shall call a message-block This new standard would allow users to enter the network at whatever data rate they wish and yet make efficient use of various data-rate channels forming the network This differs sharply from analog TELEVISION smuon i i ii i i MICROWAVE TE 91 WAVEGUIDE BURIED CABLE SATELLITE TELEPHONE LINES Fig 5 All digital network composed of mixture of links -12- transmission circuits where the bandwidth is identical all along the channel The links connecting the switching nodes operate at a very high data rate Rapid switching decisions will be necessary at each switching node to direct each pack- age message-block of bits to the next node These message-blocks must be relayed quickly from station to station so that they are formed transmitted unpacked and delivered to the recipient so as to create the illusion that a direct copper wire exists between himself and the sender HOT-POTATO ROUTING A message-block is like a letter Fig 6 In our system it consists of 866 bits of information rubber stamped with the names of the addressee and the sender together with some additional housekeeping bits forming a package of 1024 digital bits In our analogy the ad- dressee is the address on the letter the sender corresponds to the return address and the content of the enve10pe is the data being transmitted As each letter has a cancel- lation date we create an analogous quantity which we shall call the handover number Every time a message- block goes from Switching Node to Switching Node along its way this handover-number is incremented providing a measure of path length of each message-block in the network We call the routing scheme we shall use a hot-potato routing scheme because each Switching Node acts as if the message-block letter it is handling is a hot potato which must be passed on quickly before it burns the Node WHAT A MESSAGE BLOCK AN ANALocvro A LEUEP - - CANCELLATION om - maven swam mh - A- mum comma mm Memos swat moms Amount newau 71 mass 050 1 WIIHUJIUHUJHHIW acumen slam mam ans on A uus magmas _K-j l sum END OF MESSAGE sum or messnoe Miss some I sumo message 150 BITS SEC 1 540 000 message block concept -13- -14- Visualize each Switching Node containing an electronic postman at a blackboard Each postman uses the return addresses and handover-numbers of passing traffic to de- termine the apparent shortest path over each of its links to any station in the network If the best path is busy the second one is taken immediately without waiting If the second-best path is busy or destroyed the third-best path is taken and so forth Sometimes it will be neces sary even to return the message block back over the link on which it arrived Simulation shows that when each station in the network uses this simple policy an extremely effective over all switching action takes place that allows high volumes of traffic to be transmitted from any node to any other node over paths surprisingly close to being the shortest possible We find that a message-block rarely less than about one in a hundred million takes a path longer than the peripheral number of links of the network This means that with proper use of the powerful automatic error detection and repeat-transmission techniques that exist now we can maintain a very low over all error rate from user to user even though each message block may travel by different paths over noisy links THE PAYOFF FOR USING AN ADAPTIVE ROUTING DOCTRINE The simple routing doctrine described above permits the network to adapt to changes in traffic loading and link and node destruction automatically seeking the best routes in the network without human intervention To appreciate the power of self-adaptation of such a network to its environment consider the following example shown in Fig 7 In Fig 7A ABLE BAKER CHARLIE DOG EASY A Fig 7 Adaptability to change of user location -16 FOX are the locations as shown In Fig 7B ABLE moves to a different location TTmEnetwork which has learned where ABLE was soon realizes that ABLE is transmitting from a new location and now delivers messages to ABLE at its new location These adaptive routing doctrines were found to work efficiently when simulated For example in a network of 49 stations in which no node knew where the other nodes were at a time 0 it was found that within one second of scaled real world time all stations were efficiently routing traffic to all other stations This network was also examined for performance while stations were destroyed and traffic in process In summary it is felt that the very simple doctrine described is able to provide highly sophisticated operation of a network of stations THE DISTRIBUTED ADAPTIVE NETWORK A SPECIFIC SYSTEM To this point we have described some of the under- lying concepts of a new communication network using the hot-potato routing doctrine to switch blocks of data from user to user Let us consider a specific application of a communication network to provide common user services for a large number of military users in the future In particular we have examined a network of 400 Switching Nodes and 200 concentrating stations called Multiplexing Stations The number of Switching Nodes was chosen to provide a highly survivable network while the number of Multiplexing Stations provides service to at least 100 000 separate users Operating a wide variety of input devices -17- including start stop teletype machines and good quality HIDM voice telephone Each Multiplexing Station in Fig 8 is connected to three switching centers Figure 9 represents one way in which the user visualizes his communications Each user or subscriber has a terminal device such as a telephone connected to a Multiplexing Station The Multiplexing Station is similar to a PBX or telephone central office Signals from the subscriber are packaged into message-blocks and transmitted at high speed through the network of Switching Nodes terminating at the called subscriber The entire network of Switching Nodes can be considered a black box comprising a distributed transmission plant that allows delivery of message blocks from itself to the remote Multiplexing Station with minor regard for individual node failures caused by enemy action or reliability problems How the message-blocks travel is of no importance to the end users Although the entire system being described is basically a store and-forward-transmission system it can in turn be viewed by the users as a black box containing a fixed time delay of about one-half second Aside from this half- second delay the illusion is maintained that a copper-wire real-time circuit exists between the two users How are analog voice signals sent over this all-digital system The signal from the telephone microphone is con- verted into a digital signal for transmission In our net work we shall use HIDM for voice signals This is a simple digital modulation system that allows efficient digital voice signal reconversion representing a balance between Fig 8 lnterconnection between multiplexing stations and switching nodes -18- LOW SDEED TEDMINAL DATA DATE DEVICE CALLED HIGH SPEED 309550359 CALLING MESSAGE BLOCKS SUBSCRIBED TDANSMISSION CALLING MULTIDLEXING STATION 5mm 10 A IDBX7QH2 CENTDAL 31 OFFICE a an _19_ DISTDIBUTED TDANSMISSION CALLED sumou 7 Bill f view of the communication network -20 good not excellent quality low data-rate and economy of circuitry 19 200 bits per second appears to provide highly intelligible speech Push buttons on the telephone change the connections of a counter circuit within the telephone to generate a repeating binary pattern as each push button is sequentially depressed Our digital telephone contains one additional circuit that is of inter est When no voice signal is present between words or when the other party is Speaking the output signal is suppressed No message-block need be sent unless a voice sample has been heard in the last one-twentieth of a second Thus a good quality telephone conversation represents a binary load to the network of only about 5000 bits per second per digital telephone Only one person speaks at a time and about 50 per cent of the time of voice trans- mission is silence Unlike vocoder conversation systems there is no loss of naturalness of the voice yet the data rate taken to transmit the voice is comparable to a very high quality vocoder Secrecy equipment is most economically provided as an integral part of switching apparatus Both switching and processing require digital equipment that can be time-shared Figure 10A defines end-to end in which clear text is locally and transmitted to an end secure area containing a decoder The use of an identical key in the coder and decoder permit deciphering the transmitted stream Where many tandem stations are required link-by link can be used Each message originator holds a key only to his local switching center Each such SAME KEY muss TEXT I mm sun I comm 1 war 1 4'9 succvmo co Misuse 1m 1 mm END 871110 le Ammu common I I I I I I SECUQEADELJ 350363994 - omen 51mm B LINK-BY-LINK KEY ust 7m SWITCH SWTTEH STATION STATION I sun mama I mm I sewn men C DOUBLE CE ENCODEQ co cnvoro DECODEQ Fig 10 Types of systems -21- -22- switching center has equipment for its outgoing line to its adjacent Switching Node This arrangement does not require the message originator and the end ad dressee to have identical keys However as the informa- tion is available in the clear at each switching center one must have great confidence in the integrity of the personnel at each and every center Fig 10B shows a more secure arrangement It uses two layers of one for the text and another for the headings needed to direct traffic It requires less concern about security at intermediate Stations Providing separate keys between message originators and end addressees still Stands in the way of the full flexibility we would like Therefore we will design the system somewhat akin to the double tion of Fig 10C The steps of establishing a telephone call connection between two network users appear in Fig 11 As in the conventional telephone system the calling subscriber lifts his telephone off the hook indicating that service is de- sired and immediately hears the dial tone in our case the dial tone is silence The calling subscriber depresses push-buttons on his telephone This transmits the called address via the calling Multiplexing Station to the called Multiplexing Station If the called subscriber is connected his telephone rings Either a busy signal returns or an agreement to accept the call In our network we add one feature The calling Multiplexing Station transmits to the called Multiplexing Station something we call a start number In each Multiplexing Station we hold a key reserved for conversations between it and each of about CALLING CALLED MULTIPLEXING MULTIPLEXING suggl le'igm STATION STATION INSTANTANEOUSLY Wk 1mm SEND AGQEEMENT To ACCEDT CALL EITHED QETUQN BUSY SIGNAL STAQT SIMULTANEOUSLY EA EITHED DADTY 5'5 DELEASE I IS ALLOWED TO KNOW DOWN CALL 11 0peratlons performed in setting up a call between network users TIME -23 -24- 1000 Multiplexing Stations The start number cor- reSponds to the number of calls made between the calling and called Multiplexing Station that particular day If both Multiplexing Stations agree that this is in fact the 1003rd call then the digital key base assigned for this combination of two stations is modified by a second key and a new key base created to be used for the call being established Since the number of modifying opera- tions performed to the start number is large and only a small part of the possible sequence generated is ever used information from one telephone call cannot be used to help break a later call The significance of this statement is that it is safe for the network to handle mixed classified and unclassified traffic Information from one telephone call cannot be used to determine the key sequence in order to decipher the next telephone call After this exchange the two generators are in and the remainder of the call continues under control While we have shown only one level of another level on a link-by link basis is included in the Multiplexing Station and at each Switching Node to every other Switching Node Although we have discussed two levels of with keys reserved on a per-call basis our primary pro tection comes from still a third source a form of commonly described in the open literature as autokey In the autokey procedure the preceding text forms part of the key Thus it is necessary to correctly receive all the text of a message before being able to decode the next arriving symbol Autokey while being a -25- very powerful secrecy system is not practical in a net- work containing sources of error A single error destroys subsequent communication But in the network we are describing a very low error rate is expected -perhaps one error per year per user Thus we can afford to use the autokey principle for our end to-end protec- tion Let us recall that message-blocks will travel by different paths through the network A fraction of a second after a call is established the probability that all message-blocks will arrive by the same paths is very low The significance of this statement is that only the end Multiplexing Station will ever see all the message- blocks transmitted after a fraction of a second Persons tapping lines between stations can never be privy to all sequential message blocks they have a high probability of missing at least one Thus even if an enemy agent had all the keys to the network and listened from a central point he would be unsuccessful because only the end Multi- plexing Station receives all the message-blocks and autokey requires that all message-blocks be correctly received COST One skates on thin ice while describing the cost of a large system never built using new techniques But this does not absolve the system designer of the obligation to discuss costs This is ours Fig 12 The described network might be built to serve 100 000 users at an annual cost of about $60 million per year in cluding amortization These are not conservative costs as they assume that we will take advantage of the low-cost Purchasing major network ITEM QUANTITY 05590313 SWITCHING N00ES 400 150 000 00 000 000 200 300 000 00 000 000 EN0 TE0MINA1 DEVICES 100 000 200 20 000 000 LINVS MI 120 000 400 40 000 000 100 000 000 Cost summary ITEM TOTAL QESEAQCH AN0 DEVELODMENT 23 700 000 ENGINEEDING AN0 22 100 000 DUQCIIASE or M01000 NETwoczV INVESTMENT 100 000 000 INITIAL INVESTMENT $235 000 000 BASIC ANNUAL can 10 VEA0 BASIS 00 000 000 Fig 12- -Approximate cost estimate 26 -27- unreliable equipment that can be built and used in such a network But even though we will use unreliable equip ment we expect to achieve better reliability than we are accustomed to today Systems such as this can be built only if one appreciates that unit reliability and systems reliability are two separate things and that in the properly designed system system reliability can be greater than unit reliability unlike some systems that have been built Building a lot of cheap in lieu of a lesser amount of eXpensive equipment is an art the military has not really practiced for a long time SUMMARY Let us again turn to Fig 2 and compare what we have proposed against our future communication problems 1 Survivability We propose the use of an all digital distributed network using adaptive routing able to withstand heavy destruction and operate effectively after attack 2 Secrecy We propose the use of pro- visions built into the switching apparatus forming an integral part of the system We prOpose the use of schemes even more secure than the keys themselves 3 ErrorlRate Quality 'We propose the use of integral automatic error detection repeat transmission means to allow the user-to-user error rates of less than 108 bits This is several orders of magnitude better than now found in practice 4 Delay Times With the exception of the half second delay in transmission time approximately equivalent to what is noted on a high-altitude satellite transmission circuit we permit automatic user to user transmission without switching delays 28 5 Volume We propose a network with orders-of magnitude more data transmission capability than our present military communications network 6 Flexibility We propose instantaneous user-to user communications among a large number of potential users without requiring pre-arrangement of circuit assignment or keys 7 Cost While this is an expensive system the cost appears to be roughly only about six per cent of what we may now be paying for communications in the military The price may be a bargain for what it buys It might even be an economical system for certain commercial data-transmission applications Advantages and disadvantages of the distributed network concept are listed in Figs 13 and 14 respectively -29- NETWORK USES ADAPTIVE LEARNING LESS VU LNERABLE MEETS FUTURE MILITARY REQUIREMENTS HANDLES BROAD MIX OF INPUT DEVICES LARGE NUMBER OF VARIOUS TYPES OF USERS NO CUMULATIVE VOICE DISTORTION - LOW ERROR RATE CAN USE MIXTURE OF LOW COST EVEN NOISY UNRELIABLE LINKS SILENCE PERIODS SUPPRESSED ALL USERS PROTECTED HIGHLY IMMUNE TO SOPHISTICATED SABOTAGE HOLDING KEYS DOES NOT ALLOW EAVESDROPPING COST COMPARABLE TO SOFT ANALOG NETWORKS Fig 13 Advantages of distributed network -30- DIFFICULT TO EXPLAIN CONCEPT MUST UNDERSTAND COMPUTERS TO EVALUATE FEASIBILITY SYSTEM NEVER BEFORE BUILT IS EXPENSIVE TO SIMULATE LARGE NETWORK REQUIRED ONE HALF SECOND DELAY IN VOICE TRANSMISSION PERFORMANCE VULNERABLE TO POOR DESIGN LOW COST WILL HINGE ON CAREFUL DESIGN EXISTING TERMINAL DEVICES SOMEWHAT EXPENSIVE CONVERSION REQUIRED NEED FOR SURVIVABLE SECURE ERROR-FREE COMMUNICATION NOT ALWAYS APPRECIATED NOT YET OFFICIALLY REVIEWED LARGE SYSTEMS COST A LOT OF MONEY Fig 14 Disadvantages of distributed network -31- APPENDIX Further information may be found in the following publication series entitled 0n Distributed Communications 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 net- work 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 Determination of in a Distributed Network J W Smith 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 over- lapping validity -32- IV Priority Precedence and Overload Paul Baran The creation of dynamic or flexible priority and precedence structures within a communication system handling a mixture of traffic with dif ferent data rate urgency and importance levels is discussed The goal chosen is optimum utiliza- tion of the communications resource within a seriously degraded and overloaded network V History Alternative Approaches and Comparisons Paul Baran A background paper acknowledging the efforts of people in many fields working toward the develop- ment of large communications systems where system reliability and survivability are mandatory A consideration of terminology is designed to ac- quaint 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 thermo electric converter burning L-P fuel VII Tentative Engineering Specifications and Preliminary Design for a High-Data-Rate Distributed Network Switching Node Paul Baran 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 com- plex processing takes place The node is de- scribed in sufficient engineering detail to -33- estimate the components required Timing calcu- lations together with a projected implementa tion scheme provide a strong foundation for the belief that the construction and use of the node is practical 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 demonstrat- ing how the network will simultaneously process traffic from up to 1024 separate users sending a mixture of start stop teletypewriter digital voice and other signals at various rates IX Security Secrecy and Tamper-Free Considerations Paul Baran 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 protec- tion is best obtained by raising the price of espied information to a level which becomes ex- cessive 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 configura- tion 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 v$60 000 000 per year -34- XI Summary Overview Paul Baran Summarizes the system proposal highlighting the more important features Considers the particular advantages of the distrubuted network and comments on disadvantages An outline is given of the manner in which future research aimed at an actual imple- mentation of the network might be conducted To- gether with the introductory volume it provides a general description of the entire system concept This document is from the holdings of The National Security Archive Suite 701 Gelman Library The George Washington University 2130 H Street NW Washington D C 20037 Phone 202 994-7000 Fax 202 994-7005 nsarchiv@gwu edu