Preface
`
`In large measure the traditional concern of communications engineers has
`been the conveyance of voice signals. The most prominent example is the
`telephone network, in which the techniques used for transmission multiplex-
`ing and switching have been designed for voice signals. However, one of
`the many effects of computers has been the growing volume of the sort of
`traffic that flows in networks composed of user terminals, processors, and
`peripherals. The characteristics of this data trafiic and the associated perfor-
`mance requirements are quite diflerent from those of voice traffic. These
`differences, coupled with burgeoning digital technology, have engendered
`a whole new set of approaches to multiplexing and switching this traflic.
`The new techniques are the province of what has been loosely called
`computer communications networks.
`The subject of this book is the mathematical modeling and analysis of
`computer communications networks, that is to say, the multiplexing and
`switching techniques that have been developed for data traffic. The basis for
`many of the models that we shall consider is queueing theory, although a
`number of other disciplines are drawn on as well. The level at which this
`material is covered is that of a first-year graduate course. It is assumed that
`at the outset the student has had a good undergraduate course in probability
`and random processes of the sort that are more and more common among
`electrical engineering and computer science departments. (For the purpose
`0f review, but not first introduction, the required background material is
`given in a pair of appendices.) The material in the text is developed from
`this starting point. The objective is to develop in the student the ability
`to model and analyze computer communication networks. We also seek to
`impart a critical appreciation of the literature in the field.
`In a book at this level, it is inevitable that the choice of the particular
`material to be included in the text is heavily influenced by the author’s own
`research and professional experience. However, as the book evolved, the
`vii
`
`r
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 1
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 1
`
`

`

`viii
`
`Preface
`
`preface
`
`ix
`
`Jerry Foschini and John Silvester read an earlier version of the text and
`made a number of valuable suggestions for its improvement. Their dlligent
`and encouragement were
`significant contributions. Thelma
`Hyland typed the manuscript. She was the soul of patience through the
`numerous revisions. A picture is worth a thousand words, and a thousand
`thanks are due to Miao Duen-Zhi for her assistance With the figures. The
`book evolved from notes of a course taught at McGill Umversrty over the
`period 1979-1983. A number of students suffered through the earlier (and
`later) versions of the notes. For their forbearance we express our deep
`gratitude. As mentioned earlier, the material in the book reflects our research.
`Our interest in computer communications began at Bell Labs, and was
`stimulated by Bob Lucky and Jack Salz. We would also like to acknowledge
`our first collaborator in the field, Dave Sherman. Finally, the author would
`like to pay tribute to his wife, Florence, and to his children, Mary, Ann,Jemmy,
`and Martin. They endured his absence and preoccupation during the long
`writing process.
`
`JEREMIAH F. HAYES
`Concordia University
`Montreal, Quebec
`
`work of others assumed a more and more prominent role. In fact, one of
`the most rewarding aspects ofwriting the b00k was learning and aPPYeCiafing
`the fine work that others have done, and getting a sense of how the field
`has progressed. We hope that we have conveyed some of this to the reader.
`Our appreciation of a good deal of the material outside our immediate
`interests was gained to a significant degree through a number of first-rate
`survey papers in the field. We shall gratefully acknowledge these papers at
`the appropriate points in the text.
`The level of the course for which this book is intended places certain
`limitations on the material that could be covered. In several instances
`important work could not be included simply because students could not
`be expected to have the appropriate mathematical training. In other cases
`the analysis was simply too involved. In spite of these difficulties, we derive
`great satisfaction from the fact that we were able to include almost all of
`the work which we consider to be seminal in the field. We tried very hard
`to present this work in a form that would be palatable to the serious though
`inexperienced student. A number of exercises illustrating the material are
`included.
`Since the focus is on modeling and analysis, somewhat less attention
`is given to the details of implementation and operation. However, we
`attempted to include enough of this material so as to place the mathematical
`models in the proper context. In the course of the discussion we direct the
`reader to a number of authoritative sources on implementation and oper-
`ation.
`'
`There is more material in the text than can be covered in an ordinary
`graduate course. In order to assist the instructor in selecting material, we
`point out several natural groupings. Chapters 1, 2, and 3 serve as an
`introduction to the rest of the text. After completing Chapter 3 one could
`go on to Chapter 4 or to Chapter 10. The main results in Chapters 4, 5,
`and 6 are obtained through the imbedded Markov technique, thereby
`providing a unifying theme for the three chapters. The results in Chapters
`6, 7, 8, and 9 are pertinent to the modeling of a particular form of computer
`communication network—the local area network. In fact, all of the well-
`known accessing techniques are considered. In order to cover Chapters 6
`through 9, some material on the imbedded Markov chain is also necessary.
`Chapters 10, 11, and 12 also provide a convenient grouping. Chapters
`10 and 11 rely heavily on Jackson network theory. The linkage between
`Chapters 11 and 12 is that they treat the so-called “higher-level" protocols,
`in particular flow control and routing. Although the final chapter on network
`layout stands somewhat apart from the others, there are points of contact
`between network layout and routing. This latter is the subject of Chapter 12.
`In concluding these opening remarks we would like to express our
`
`deep gratitude to a number of people for their contributions to this work. effort
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 2
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 2
`
`

`

`Contents
`
`Chapter 1. Basic Orientation .....................................
`Ll. Modeling and Analysis of Computer Communication Networks .
`.
`. .
`1.2. Computer Communications Networks ...........................
`1.3. Summary of Text .............................................
`
`Chapter 2. Protocols and Facilities ...............................
`2.1. Introduction .................................................
`2.2. Properties of Data Traffic .....................................
`2.3. Protocols ....................................................
`2.4. The Telephone Network ......................................
`2.5. Data Transmission ...........................................
`2.6. Data Networks ...............................................
`2.7. Local Area Networks .........................................
`2.8. Alternative Facilities: Satellites and Cable Television .............
`
`>
`
`Chapter 3. Pure Birth and Birth-Death Processes: Applications to
`Queueing ..... . ................. I......................
`Introduction ................................................
`3.1.
`3.2. Bernoulli Trials—Markov Chains ..............................
`3.3. The Poisson Process .........................................
`3.4. Pure Birth Processes .........................................
`3.5. Birth—Death Processes .......................................
`3.6. Queueing Models ...........................................
`3.7. Delay—Little’s Formula ......................................
`38. Burke’s Theorem ............................................
`3-9 Communications Example ....................................
`3-10. Method of Stages ...........................................
`Exercises
`
`mm.—v-I
`
`10
`ll
`14
`21
`31
`34
`42
`
`49
`49
`49
`50 '
`57
`59
`62
`71
`77
`80
`80
`85
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 3
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 3
`
`

`

`Chapter 2
`
`PI,mocols and Facilities
`
`35
`
`A F
`
`igure 2.20. (A)—(D) Ring, bus, star, and tree configurations.
`
`The ring topology is a sequence of point-to—point links with flow in
`one direction around the ring. As we shall see, there are several acceSsing
`techniques for the ring system. In all of these there is a delay due to
`processing at each of the stations. For reasons of reliability, there are
`Provisions to bypass stations if they become inoperative.
`In both the bus and the ring tOpologies the control of traflic is dis-
`tributed. A topology in which control is concentrated is the star. In this
`Case all of the traffic in the system is switched in a central “hub.” The .
`Stations access the hub through high-speed lines.
`A fourth topology, the tree, has been used to distribute data over a
`Wider area. Typically, in networks using the tree topology, remote stations
`aCcess a central processor. A ubiquitous example of such a network uses
`multipoint private lines in common carrier networks. The tree topology
`Would also be relevant for CATV networks used to transmit data, since
`these sorts of networks are closely related to LANs and the same modeling
`teChniques are relevant.
`
`
`
`34
`
`2.7. Local Area Networks
`
`2.7.1. General Considerations
`
`The second major class of networks that we shall consider are local
`area networks (LANs), which are data networks having limited geographical
`area, usually within a kilometem‘ Networks confined to a single office
`building, shopping center, or university campus are prime examples of
`LANs. The emergence of this sort of network is part of the general growth
`of computer and digital technology; however, the introduction of office,
`automation and distributed processing systems has furnished additional
`impetus. In both of these applications LAN techniques play a significant
`part.
`
`In the past, local area networks were defined in terms of geographical
`extent and data rate; however, in view of the rapid growth of the technology,
`a more useful definition may be in terms of usage and configuration. In
`providing a common channel for a number of users in a limited geographical
`area the emphasis is upon ease and flexibility in providing access. Due to
`the limited geographical area, bandwidth is not the critical commodity that
`' it is in networks covering a wide area. Therefore, access to the network can
`be simplified at the cost of bandwidth. A second point is that data networks
`covering a large area require redundancy with respect to connectivity in
`order to ensure operation in the face of failures. For example, the ARPA
`net requires at least two paths between source—destination pairs. Because
`of the limited geographical extent of the typical LAN, it is in something of
`a protected environment and this sort of redundancy is unnecessary. This
`simplifies the topology since only a single path need be provided between
`source—destination pairs.
`
`2.7.2. Topology
`
`In current practice three basic topologies are prevalent in local area
`networks: the bus, the ring, and the star (see Figures 2.20a—2.20c). A fourth
`topology, the tree (see Figure 2.20d), is used in related systems and may be
`considered as a form of the bus configuration.
`The bus topology is appropriate to transmission media such as coaxial
`cable or radio which allow what are, in effect, high-impedance taps. In
`principle, these taps do not afiect the medium, and a large number of
`stations can be connected. Each of these stations can broadcast simul-
`taneously to the others. The bus topology is particularly appropriate for
`the random accessing techniques that we shall be discussing presently.
`
`TSurveys of LAN techniques are contained in Refs. 41—43. See also Ref. 44.
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 4
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 4
`
`

`

`36
`
`Chapter 2
`
`protocols and Facilities
`
`The four topologies can be the basic building blocks for more complex
`topologies. For example, a system of interconnected rings has been pro
`posed. In the current standard involving the bus topology a series of
`connected buses form a kind of rootless tree.
`
`2.7.3. Transmission Media
`
`In the foregoing discussion of topology we alluded to the transmission
`media used to implement the network. In current practice three kinds of
`transmission media are used in LANs: twisted pairs, coaxial cable, and
`optical fiber. The twisted pairs of copper wires are the same as those used
`in the telephone plant. Because of the readily available technology, there
`is a strong tendency to operate at the same rates as short-haul digital carrier
`systems, i.e., the T1 rate 1.544 Mbps. Higher speeds are possible as well.
`The nature of twisted pairs is such that it is basically a point-to-point
`medium. As mentioned earlier, coaxial cable can operate as a multiple
`access medium by use of high impedance taps. Rates of up to 10 Mbps are
`attained in commercial systems. Optical fiber is the medium with the highest
`transmission rate currently used in LANs. Data can be transmitted at rates
`of up to 50 Mbps using light emitting diodes. With Laser transmitters, rates
`in the range of hundreds of Mbps can be achieved without repeaters over
`distances compatable with LAN operation. At the present writing optical
`fiber is basically a point-to-point medium. However, the development of a
`low-loss optical tap could change this situation.
`In the future other transmission media may play a role in LAN and
`related systems. Radio, for example, would allow more flexible operation
`than the media we have considered. A radio medium would allow a random
`accessing technique to be employed. A second widely available medium
`which could be adapted for two-way data transmission is the CATV network.
`We shall be considering this medium in some detail in the sequel.
`
`2.7.4. Access Protocols
`
`The fundamental purpose of the local area network is to allow data
`sources that are dispersed throughout an area to share a common trans-
`mission medium. Because of this dispersal, transmission capacity must be
`expended to coordinate the flow of traflic from each of the sources. The
`way that this is done is the function of the access protocol. A significant
`portion of this book is devoted to the modeling and analysis of these access
`protocols. The objective of the analysis is to evaluate performance in terms
`of delay, storage, and throughput.
`
`2.7.4.1. Polling
`The predecessors to the current local area networks were tree networksiri
`' h provide communications between a number of remote stations an,
`Whlc
`3,1
`rocessor. Such an arrangement would be appropriate to a bank 5
`a “I?“ hgcking system, for example. The coordination of traffic from the
`credit 6 ' efiected by roll call polling. Each station is assigned an address.
`station
`5 IS
`0115 stations for messages by broadcasting the
`The central processor p
`e reception by a station of its address is, in effect,
`.
`addresses in sequence. Th
`on has a message to transmit, it interrupts
`a license to transmit. If a stati
`the polling cycle.
`eled and analyzed in Chapter 7 of the text.
`Polling systems are mod .
`.
`he eflect of overhead on performance.
`The salient result of this anainlSJS t
`overhead in this case is the time required to poll all of the stations even if
`none of the stations has a message to transmit. In many atpplicatiogiiiaa
`significant portion ofthe overhead is due to the establishment o coanu e of
`tions between the remote stations and the central processor. In t 3 cas .
`voiceband modems, for example, this would require phase an
`timing
`recovery and equalizer training for the higher speeds.
`An alternate technique, called hub polling, attempts to reduce this
`overhead. In this case the license to transmit is transferred between stations
`directly without going through a central processor. This technique is similar
`to those used in ring systems, which we shall be discussmg in the next section.
`
`2. 7. 4.2. Ring Protocols
`From the perspective of current techniques, the first local areanetwork
`was the ring system built by Farmer and Newhallfm This system is shown
`in Figure 2.21.: The flow of data around the ring is organized into the frame
`“fl Chapter 13 we shall be considering the layout of tree networks.
`1A survey of LANs with the ring topology Is given in Ref. 46.
`
`
`
`Figure 2.21, Farmer—Newhall loop.
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 5
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 5
`
`

`

`38
`
`Chapter 2
`
`.
`Is “4 Flcllitm
`
`39
`
`structure shown in Figure 2.21. At any given point in time only one station
`on the ring is authorized to transmit. Message transmission begins with a
`start of message (SOM) character. This is followed by source and destination
`addresses and by the data to be transmitted. Stations along the ring monitor-
`the line flow and read messages which are addressed to them. The transmit.
`ting station relinquishes control by appending an end of message (EOM)
`character to the data. Upon receipt of the EOM character the next station,
`downstream has the opportunity to take control of the ring. If this station
`has nothing to transmit, the EOM character is passed on to the next station,
`and so the cycle goes. In later implementations of the ring the term “token"
`is used to indicate control. Again the bearer of the token has control which
`he relinquishes by transmitting it. A station with a message to transmit
`seizes control when it observes a free token on the line. In this case the
`token can simply be a one-bit field at the beginning of a frame. The
`mathematical models of the Farmer—Newhall and the token passing rings
`are basically the same polling model that we have just discussed. In all
`three systems opportunity to transmit a message is given to a single station
`at a time. Moreover, this opportunity cycles through all of the stations. For
`polling systems the opportunity consists of the poll, which, in effect, asks
`a station “Do you have a message?” In the case of the ring protocols it is
`the reception of an EOM character or a token.
`A drawback to the token passing technique is that only one station
`may transmit at a time. If traflic is symmetric around the loop, this means
`that only half of the capacity is used on the average. If bandwidth is not
`scarce in LANs then this may be an advantageous trade for simplicity of
`operation. However, there are alternative techniques which allow more than
`one transmitting station at a time. One of these is what is called demand
`multiplexing and is illustrated in Figure 2.22.‘47"8) Flow on the line is
`segmented into fixed size frames. At the beginning of each of the frames is
`a one-bit field indicating whether or not the frame is occupied. Stations
`having data to transmit seize empty frames as they pass. The indicator bit
`is changed from empty to full and addressing information is transmitted
`along with the data. Demand multiplexing requires that each frame of
`packet have addressing, while for token passing a single address would
`suflice for an entire multipacket message. The demand multiplexing tech-
`nique is analyzed by means of priority queues in Chapter 6 of the text.
`Priority queues are appropriate in this context since traflic already on the
`line has priority over locally generated traflic. Furthermore, the priority is
`what is called preemptive inasmuch as the transmission of messages consist-
`ing of more than one packet can be interrupted by packets already on the
`line.
`
`once the transmission of a messages has begun. Messages arriving on the
`
`A final alternative, called bufler insertion, does not allow interruption
`
` - ATA SLOT
`
`INDICATOR
`
`BIT
`
`Figure 2.22. Demand multiplexing.
`
`line from other stations are buffered while the locally generated message is
`being transmitted. For bufler insertion messages are ‘kept intact and more
`than one station may transmit at a time. The bufier insertion technique is
`anal
`ed in Cha ter 6.
`.
`also In egh of the rililg accessing techniques the idea is to cope With the
`burstines of data sources by giving line access only as it is required: It is
`useful to compare this kind of technique to one in which transmisswn
`capacity is dedicated to particular data sources. In (Fhapter 5 we study the
`performance of time-division multiplexing (TDM), in which time slots are
`dedicated to data sources sharing the same line.’r
`
`2.7.4.3. Random Access Protocols
`
`Random access techniques first developed for radio systems‘”’5°’ have
`been successfully applied to local area networks. The origin of these tech-
`niques is the ALOHA protocol, which is a form of completely distributed
`control. Again it is assumed that several stations are sharing the same
`transmission line. Furthermore it is assumed that the line is of the bus
`cutlfiguration (see Figure 2.23). When a station generates ahxed-length
`packet, it is immediately transmitted without coordinating With the other
`stations. As usual, the packet contains addressing information and parity
`Check bits. If a message is correctly received by a central controller, a
`positive acknowledgment is returned to the transmitter. Since there is no
`coOrdination among the stations, it may happen that two or more stations
`interfere with one another by transmitting at the same time. If messages
`
`TTDM was discussed in connection with the T1 digital carrier system. In the context of a local
`area network, the TDM concept is called time-division multiple access (TDMA).
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 6
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 6
`
`

`

`40
`
`COLLISION
`
`SUCCESSFUL
`TRANSMISSIONS-
`
`Chapter 2
`
`Protocols and Facilities
`
`41
`
`p
`10‘
`2°‘I:E:J
`
`/\
`[I] D
`
`N 0'
`
`Figure 2.23. The ALOHA technique.
`
`collide, the resulting errors are detected by the central controller, which
`returns either a negative acknowledgment or nothing. An alternate
`implementation is to have the transmitting station monitor the line for
`collisions. After a suitable time-out interval a station involved in a collision
`retransmits the packet. In order to avoid repeated collisions, the time-out
`intervals are chosen at random independently at each station. The analysis
`of the ALOHA technique is presented in Chapter 8. The results show that,
`because of collisions and the resulting retransmissions, the throughput is
`no more than 18% of the line capacity. Moreover, there is a basic instability
`in the system.
`,
`The basic ALOHA technique can be improved by rudimentary coordi-
`nation among the stations. Suppose that a sequence of synchronization pulses
`is broadcast to all stations. Again assume that the stations generate fixed-
`Iength packets. The interval between synch pulses is called a slot and is
`equal to the time required to transmit a packet. Packets, either new or
`retransmitted, can only be transmitted at a pulse time. This simple device
`reduces the rate of collisions by half since only packets generated in the
`same interval interfere with one another. In pure ALOHA this “collision
`window” is two slot intervals. This modification is called slotted ALOHA?”
`In Chapter 8 it is shown that there is saturation at approximately 36% of
`capacity. However, in spite of the increase in throughput, the unstable
`behavior persists.
`An extension of ALOHA that is particularly apprOpriate for local area
`networks is carrier sense multiple access (CSMA).‘52) Before transmitting a
`newly generated message, a station listens on the common channel for the
`signal of another station. If the channel is free, the message is transmitted;
`if not, transmission is deferred. Because of delay in the system, collisions
`can still occur. As in the basic ALOHA technique, conflict is resolved by
`random retransmission. Variations in conflict resolution techniques involve
`the retransmission strategy. The results of anlysis show that for local area
`networks sensing potential conflict considerably improves performance. The
`results also indicate the same sort ofinstability that is present in the ALOHA
`systems.
`
`
`
`2 7,4,4. Tree Search Techniques
`
`The deleterious eflect of overhead on polling systems can be lessened
`by means of an adaptive tree search technique which has been given the
`name probing. A similar technique can be used to eliminate instability 1n
`random access systems. The technique is modeled and inalyzecl in Chapter 9.
`In the application of the technique to polling systems} )a basrc assumption
`is that a central processor can broadcast to all stations Simultaneously. The
`essense of the technique is to poll stations 1n groups rather than one at a
`time. If a member of a group being polled has a message 'to transnnt 1t
`responds by putting a signal on the line. Upon receivmg a posrtive response,
`the central processor splits the group in two and polls each subgroup in
`turn. The process of splitting groups continues until stations havmg messages
`are isolated, whereupon the messages are transmitted. Clearly we have a
`form of tree search. As the process unfolds the probability of statiOns having
`messages varies. Accordingly, the sizes of the initial groups to be polled
`are adapted to the probability of their having messages to transmlt. The
`criterion for choosing the group sizes is the minimization ofthe time requrred
`to grant access to all of the stations. Comparison of the technique with
`conventional polling shows a considerable improvement in performance at
`light loading. Moreover, due to the adaptivity there is no penalty at heavy
`loading.
`.
`As mentioned above, the tree search technique is also appropriate to
`random access systems. Suppose that in response to a signal from a central
`processor, a station simply transmits any message that it might be harboring.
`Conflicts between stations responding to the same signal are detected by
`the processor. In order to resolve the conflict the group is split in two and
`the subgroups are polled. The process continues until all of the subgroups
`have only one station with a message to transmit. As in polling systems,
`Optimal initial subgroup sizes can be chosen so as to minimize the time
`required to read message out of a group of stations. Optimum group sizes
`are chosen adaptively as the process unfolds.
`Control of the adaptive process need not be as centralized as the
`fOTCEOing impliesf”) As in slotted ALOHA, the transmission process can
`be Synchronized by a sequence of pulses broadcast to all stations. By explicit
`or inlplicit feedback, stations are informed of successful transmissions and
`0f collisions. This information determines the response of an individual
`“mm to a synchronizing pulse. As in the polling case there is an adaptive
`algorithm which minimizes the time required to grant access to all stations
`having messages to transmit. Analysis of the adaptive technique shows
`that its thoughput is 43% of capacity. This contrasts with slotted ALOHA
`Where the saturation point is 36% of capacity. A possibly greater advan-
`tage is that there are no unstable states where the system is saturated by
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 7
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 7
`
`

`

`42
`
`Chapter 2
`
`"$06015 and Facilities
`
`43
`
`retransmissions and conflicts. In the case of heavy loading each station is
`aSSigned a slot and the system reverts to TDMA.
`
`2.7.4.5. Reservation Systems
`
`Related to random access multiplexing are a number of techniques in
`which stations reserve part of the capacity of a channel when they have
`messages to transmit.(55’55’ The reservation techniques are appropriate to
`sources which are active infrequently but which transmit a steady stream
`when active. Trafiic from such sources is not bursty; however, the requests
`for service can be treated as bursty traflic. The random access and the
`adaptive techniques discussed above would then be appropriate.
`
`2.7.5. Optical Fiber Systems
`
`Optical fiber is suited to local area networks because of the high data
`rates that it can sustain over the appropriate distances. Moreover, these
`rates are achieved at relatively low cost. Optical fiber has other advantages
`for LANs. A typical fiber together with cladding is small and flexible and
`consequently easy to install. Secondly optical fiber is immune to electromag-
`netic interference. The network topology most often associated with optical
`fiber based systems is the star,” although a ring based system has been
`deployed as well. In- both the ring and the star systems the fiber is used as
`a point-to-point medium. A factor that militates toward a star configuration
`is reliability. By its nature, fiber is nonconducting. Thus, it is not possible
`to power equipment at a user station through the transmission medium. It
`is therefore advisable to concentrate essential equipment at one place, as
`in the star. A second factor that may come into play is processing power.
`In order to take advantage of the high speed of the fiber, processing speeds
`should be high. At the speeds with which we are dealing, concentrating
`computation in one place seems to be the better approach. At this writing
`optical fiber LANs are at an embryonic stage of development. In the two
`star systems that have been discussed in the literature, the hub acts as
`reflector.
`
`end service. In both of these systems there is the means to bring high data
`
`2.8. Alternative Facilities: Satellites and Cable Television
`
`At the present writing the vast bulk of data trafl'ic is conveyed by the
`telephone networks including value added networks such as TYMNET and
`TELENET. However, the advance of technology in the form of satellites and
`cable television provides alternative data transmission facilities for end—to-
`
`rates directly to the user. Moreover, the service promises to be ubiquitous
`since the basic facilities are either already in place or can be easily installed.
`
`23,]. Satellite Communications'l’
`The commercial use of satellites for telecommunication began in 1965
`with the launching of INTELSAT 1 which provided transatlantic service. Since
`1971, when INTELSAT IV(58) was launched, satellites have been a basic
`component of transoceanic transmission systems. This system provided 6000
`voice channels. Later INTELSAT systems provide as many as 18,000 voice
`circuits. The first domestic system was the Canadian TELSAT which went
`into service in 1973, since then there has been a rapid sucession of domestic
`and international communications satellites.
`Unlike some of its experimental predecessors, the modern communica-
`tions satellite is active in that received signals are amplified and retransmit-
`ted. In early experimental systems the satellite was simply a passive reflecter.
`Another standard feature of modern communications satellites is that they
`are in the geostationary orbit at a distance of 37,000 km from the earth. The
`geostationary orbit simplifies tracking and ensures that the satellite is always
`in view. The disadvantage of this orbit is that the link delay of 0.23 sec
`afiects the way in which data may be transferred reliably. For example, this
`large delay considerably reduces the throughput of the ARQ technique,
`which is standard on domestic links.
`The three frequency bands that have been allocated to communications
`systems are shown in Table 2.1. For currently operating systems the bulk
`of the trafiic is carried over the 6/ 4 band. In the 14/ 12 hand the bands 11.7
`to 12.2 MHz are designated for domestic use while the remaining bands are
`for international trafiic.
`A number of links or channels can be established through the same
`satellite. This can be accomplished in several ways. For space-division
`TThe material in this section has been distilled from Refs. 15—17.
`
`Table 2.1
`\*;_—_
`Transmit
`Receive
`Band
`Bandwidth
`(downlink)
`(uplink)
`
`designation
`(MHz)
`(GHz)
`(GHZ)
`
`6/4
`[4/ 12
`
`500
`250—500
`
`5325—6425
`14.0—14.5
`
`3.7—4.2
`11.7—12.2
`10.95—1 1.2
`l l .45—1 1.7
`
`17.7—21.22500—350029/19 27.5—31.0w“
`
`
`
`
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 8
`
`Petitioner Cisco Systems - Exhibit 1010 - Page 8
`
`

`

`44
`
`Chapter 2
`
`protocols and Facilities
`
`45
`
`multiple access (SDMA) each link is served by a diflerent antenna. In order
`to provide for several distinct links or channels the antennas must have
`narrow beam width, hence large size. A second approach is frequency-division
`multiple access (FDMA). In this case each distinct channel between earth
`stations occupies a different frequency band. The technique is essentially
`the same as frequency-division multiplexing (FDM), which we have con-
`sidered earlier (see Section