`
`
`
`
`ACCESS PROTOCOLS
`
`.c. nETNAoHAs
`
`
`
`Protocol for accessing satellites
`efficiently: a tutorial.
`‘
`
`ATELLITES provide a convenient medium for
`data communication between widespread geo-
`graphic areas. Compared to a terrestrial data
`network, the satellite system has wide bandwidth,
`high accuracy of
`transmission, and high availability of
`transmission medium. The main disadvantages of the satellite
`system are the inherently long transmission delays (270 ms
`one way,
`the- effect of
`local weather conditions and
`interferences, and the high cost of the system. Technological
`advances can reduce the cost and the effect of weather
`
`long
`conditions on the transmitted signal. The effect of
`transmission delay can be minimized by using effective
`transmission protocols. Because of the above advantages,
`satellite technology has aroused a great deal of interest in
`recent years [7],[14].
`This paper presents a tutorial on the various protocols used
`in satellite data transmission. The most important character-
`istic of the satellite system is the ability of the earth stations,
`located at geographically dispersed areas,
`to access the
`satellite to transmit and to receive data. The area covered by
`a geostationary satellite is a function of the satellite’s receiving
`and transmitting antenna(s). For a large number of users with
`bursty traffic, a highly efficient way of using the channel
`capacity is to use multiple access techniques. In a multiple
`accessed channel, two or more users may nominally share the
`channel. The satellite system can provide broadcast capa-
`bility at any given time to all earth stations within its
`transmission coverage area. The combination of multiple
`access and broadcast capability makes
`it possible to
`configure the earth stations into a fully connected “one-hop”
`network.
`
`CHANNEL DERIVATION
`
`There are three ways to obtain channels in a satellite
`system [8]. In the first method, the channels are obtained by
`using the built-in satellite channelization due to the use of
`multiple transponders operating in different frequency bands.
`
`The author is with the Department of Quantitative and Information Sci-
`ence, Western Illinois University, Macomb, IL 61455.
`
`
`
`Each one of the independent transponders in the satellite is
`designed to accept transmission at a selected frequency band,
`i.e.,
`the uplink frequency. The satellite carries out a
`frequency translation to a well-defined frequency band, i.e.,
`the downlink frequency. This scheme thus divides the total
`bandwidth of the satellite into well-defined channels. The
`
`advantages provided by this scheme are reduced interference
`problems and improved reliability in that the possibility of
`losing all the channels due to satellite failure is small.
`The second method uses
`the basic multiple access
`techniques of frequency-division multiple access (FDMA),
`time-division multiple access (TDMA), and code-division
`multiple access (CDMA).
`
`A numberof multiple access
`protocols are presented. In the final
`analysis, it is cost which will dictate
`which protocol is suitable for
`a particular application.
`
`A simple form of obtaining an FDMA channel is to divide
`the bandwidth of a transponder into separate nonoverlapping
`subchannels, with each user assigned a separate subchannel.
`In FDMA, each user has access to a dedicated portion of the
`channel at all times. The main advantages of FDMA are that
`it
`is simple to implement in that no real-time coordination 7
`among transmission of data is needed and can be used to
`transmit either analog or digital signals. For bursty traffic, the
`channel utilization is poor. This scheme is cost effective for
`applications that involve point-to-point trunking.
`In TDMA, each user is scheduled to transmit in short
`nonoverlapping intervals. Therefore, a TDMA scheme
`requires some form of frame structure and a global timing
`mechanism to achieve nonoverlapping transmission. For this
`reason, a TDMA system is more complex to implement than
`an FDMA. However, an important advantage is the connec-
`tivity. This is obtained because all receivers listen on the same
`channel, while all sources in a TDMA system transmit on the
`same common channel at different times.
`
`The third method uses dynamic sharing of a channel using
`demand assignment techniquesfl‘his method may be used for
`circuit-switched voice traffic or packet-switched data traffic
`
`0163-6804/80/0900-0016 $00.75 © 1980 IEEE
`
`16
`
` IEEE COMMUNICATIONS MAGAZINE
`
`
`
`SATELLITE MULTIPLE
`
`
`
`
`
`APPLE 1009
`
`1
`
`APPLE 1009
`
`
`
`,_._
`SEPTEMBER 1980
`
`
`
`
`
`[8]. In this paper, we confine ourselves to packetuswitched
`data traffic only. The packet-switched data traffic system can
`be divided into random access,
`implicit reservation, and
`explicit reservation systems. In the following discussion, we
`will assume that the whole' of a transponder bandwidth is
`devoted to multiaccess operation,
`the up channel at one
`frequency operating in multiaccess mode and the down
`channel at another frequency operating in broadcast mode.
`The earth stations which are visible to the satellite antenna
`
`transmit packets atthe full available bandwidth. The satellite
`after frequency translation retransmits packets at the full
`available bandwidth. The downward packets are received at
`all earth stations within the satellite’s coverage. The earth
`stations identify packets destined to them by looking at the
`packet header address. All packets addressed to other
`stations are ignored and those addressed to the station are
`passed on to it.
`
`RANDOM ACCESS SYSTEM
`
`_One of the protocols used for transmitting packets in a
`random access satellite system is the ALOHA protocol. In
`this protocol, each one of the earth stations transmit packets
`as soon as each one of them has a packet to transmit without
`regard for other stations. Due to the lack of coordination
`among the distributed ground stations, packets from different
`stations may reach the satellite at the same time and collide,
`thereby destroying the information content. Therefore, a
`subsequent retransmission of the packet is required. Because
`of the broadcast capability posed by the down channel, the
`transmitting station will be able to detect any collision. No
`acknowledgment is necessary in the satellite system in the
`event of collision. The collided packets are retransmitted after
`a further random delay in order to avoid the risk of repeated
`collisions. The maximum channel. capacity that is usable is
`about 18 percent in the ALOHA protocol.
`A substantial increase in usable channel capacity can be
`obtained by using the S-ALOHA (slotted-ALOHA) proto-
`col. In the S-ALOHA protocol, the satellite channel is slotted
`into segments whose duration is exactly . equal
`to the
`transmission time of a single packet (assuming fixed size
`packets). If the earth stations are synchronized to start the
`transmission of packets at the beginning of a slot, the channel
`utilization efficiency increases. In the ALOHA protocol,
`when a collision takes place, the packets may overlap fully
`or partially. By using the S-ALOHA protocol, the partial
`overlap is eliminated. Under certain assumptions about the
`message traffic generated by the earth stations, the channel
`utilization efficiency is about 36 percent [1],[9]. This increase
`in channel utilization efficiency is obtained at the cost of
`increased complexity in control compared to the ALOHA
`system.
`One of the drawbacks of therandom access system is the
`problem of instability. When large numbers of stations are
`active, excessive traffic leads to more collision. After
`collision,
`the channel
`traffic consists of both the newly
`generated packets and the retransmitted packets. As the
`number of newly generated packets increase, the chance of
`
`
`
`
`in turn, increases the number of
`collision increases. This,
`retransmissions which,
`in turn, increases the chance of a
`collision, and. a runaway effect occurs; thus, the channel 1
`becomes unstable. In the absence of a control mechanism
`
`the collision retransmission may produce a
`[5],[IO],[13],
`congested condition with the system throughput becoming
`zero. The purpose of the control is to prevent the channel from
`reaching the unstable condition, while optimizing channel
`efficiency and performance during normal operating condi-
`tions [13].
`The low bandwidth utilization of the ALOHA and the S-
`
`ALOHA systems have led to many proposals for increasing
`utilization by means of slot reservation schemes. The object
`of slot reservation schemes is to reserve a particular time slot
`for a given station. This ensures that no collision will take
`place. In general,
`it may be possible to achieve potentially
`high channel efficiency using some form of a reservation
`technique. This increase in channel utilization efficiency is
`obtained at some overhead cost, either in terms of allocation
`of part of the bandwidth for reservation purposes and/ or
`increased complexity of the control mechanisms in transmit-
`ting stations. All reservations methods use some form of
`framing approach, and the reservation scheme can be either
`implicit or explicit.
`
`IMPLICI'I' RESERVATION
`The implicit reservation protocol uses a frame concept to
`the S-ALOHA- channel to permit implicit reservation. A
`frame may consist of more than one slot. The total number of
`slots can be grouped into a set of reserved slots and a set of
`slots which can be accessed using the S-ALOHA contention
`protocol. Efficient channel utilization is obtained by allowing
`stations with high traffic rate access to one or more slots from
`the reserved set of slots in each frame.
`this principle with
`The reservation-ALOHA utilizes
`implicit reservation-by-use allocation. Reservation-ALOHA
`uses distributed control, and each earth station executes an
`identical allocation algorithm based on the global information
`available from the channel. Whenever a station successfully
`transmits in a slot, all the stations internally assign that slot in
`subsequent frames for exclusive use by the successful station.
`Thus, each station maintains a history of usage of each
`channel slot for one frame duration. This slot is reserved to
`this station until the station is finished using it. The stations
`use the S-ALOHA contention method to access the
`
`unassigned slots in each frame. In this scheme, there is no
`way to prevent a station from successfully capturing most or
`all of the slots in a frame for an indefinite time period.
`
`EXPLICIT RESERVATION
`
`These reservation schemes try to make better use of the
`channel bandwidth by explicitly reserving future channel time
`for transmitting one or more messages for a specific station.
`To obtain good performance,
`the ground stations should
`cooperate with one another to maintain synchronism. Only
`by conforming to the reservation discipline can the earth
`stations ensure that packet collisions will either be eliminated
`
` 17
`
`
`
`2
`
`
`
`
`IEEE COMMUNICATIONS MAGAZINE
`I7
`
`
`
`or reduced drastically. In the explicit reservation scheme, the
`earth stations use part of the channel bandwidth for sending
`reservations for future time slots. This,
`to some extent,
`reduces the total bandwidth available for data transmission.
`
`By keeping the bandwidth required for reservation propor-
`tionately small compared to that available for data transmis-
`sion, highchannel utilization efficiency can be-obtained.
`Compared to nonreservation schemes, more complex control
`mechanisms are needed in the earth stations. The reserva-
`
`tions may be sent in separate time slots which are distinct from
`the time slots used for data transmission or they may be
`combined with data transmission (piggybacked) or both. The
`control technique used to allocate the reserved time slot may
`be central control, distributed control, or_ a combination of
`both.
`
`RESERVATION ALOHA
`
`scheme makes use of separate time slots for
`This
`reservation, with the control function distributed in all the
`stations. The satellite channel is divided into time slots of fixed
`size [ 1 1]. Every M + 1th slotis subdivided into V small slots
`as shownin Fig.1.
`
`
`
`
`Fig. 1. Satelllte channel tor reservatlon ALOHA.
`
`The V small slots are used by all the active stations to send
`reservations for future time slots and acknowledgments.
`These V small slots are accessed using the ALOHA
`contention technique. The M large slots carry reserved data
`packets. ,
`Whenever a station receives data packets to transmit, it
`randomly selects one of
`the V slots and transmits its
`reservation. This reservation is heard by all the stations. The
`distributed control
`in each of the earth stations adds the
`broadcasted reservation to the existing reservation count.
`Effectively, all the waiting packets for which a reservation has
`been made join one “queue in the sky,” the length of which is
`known at all times to all ground stations. The number of
`reserved data slots that can be reserved in one request range
`from one to eight. The requesting station has now successfully
`reserved a sequence of future time slots for data transmission.
`Once a reservation is made, each one of the stations knows
`which future slots belong to them, and no other station need
`concern itself with the details of reservations made by other
`stations.
`Fig. 2 shows an example taken from [11] which illustrates
`how this reservation scheme functions. Let us assume that the
`total roundtrip delay for signal travel is 10 slot time and there
`are five data slots (M) and six small slots (V). If a station
`transmits a reservation for three future data slots so as to fall
`
`
`
` llllllllll
`
` llIllllllllllllllllllIll
`
`
`
`Flg. 2. Satelllte channel for reservatlon.
`
`in a small slot (ALOHA) at t = 5, then all stations will receive
`this reservation request at! = 10 (the roundtrip delay). If no
`collision has taken place, then the future data slots that can be
`used for data transmission are easily calculated, provided the
`current queue length is known. Assuming the current queue
`length to be 13,
`then the station which requested the
`reservation has to wait until 13 data slots have passed by
`before it can transmit data. In our example, the slots are at
`timet = 21, 22, and 24, 23 being the ALOHA slot. Because
`it takes 5 slot time for the data packetsto reach the satellite,
`the ground station starts transmission at t = 16, 17, and 19.
`The performance of the system is a function of the value M,
`the number of data slots available between each reservation
`slot. Assuming that there are N ground stations, and if each
`one of them is allowed to reserve 'up to eight slots,
`the
`maximum allowed, then some reservations may carry over
`beyond the next reservation sIot if 8N is greater'than M. The
`system becomes overloaded if each station is allowed to
`reserve eight further slots. This increases the queue length of
`future reservations, thereby increasing packet delay. This
`situation can be avoided if each ground station is constrained
`to a limit of eight future reserved slots at any time [4].
`Another factor which may degrade the performance of the
`system is excessive contention for reservation slots. The
`number of V slots must be related to the number of earth
`
`stations and to the likely traffic activity to be expected.
`In this scheme, the channel may be in any one of the two
`states called the ALOHA and RESERVED states. On start
`up and when it
`so happens that no reservations are
`outstanding, the channel is in ALOHA state. In ALOHA
`state, the channel consists of only slots of type V. It is possible
`to send acknowledgments, reservations, and even data which
`will fit into the small slots. In this state, a reservation request
`may be transmitted in any of
`the small slots, with no
`. requirement to wait for up to M data slots to pass by. Once a
`successful reservation has been established,
`the channel
`enters the reservation state and any further reservation can be
`made in the small slot. Once again the channel enters the
`ALOHA mode if the number of reservations goes to zero.
`
`R-TDMA
`
`This explicit reservation protocol is a modified version of
`the contention and fixed assignment reservation method used
`in [2]. This scheme uses a fixed-assignment technique for
`
`18
`
`
`3
`
`
`
`
`
`‘___
`SEPTEMBER 1980
`
`
`
`making reservations and allows the total available channel
`capacity to be shared among all stations that are busy [14].
`Fig. 3 shows the R-TDMA channel. One routing frame on
`the channel is divided into a number of reservation frames.
`The reservation frame consists’of a set of reservation slots
`and a number 'of fixed length data slots. These data slots are
`grouped together to form a data frame. A reservation frame
`may have one or more data frames. Each station is assigned a
`fixed slot in each reservation frame. Each of the stations is
`assigned a fixed slot
`in each one of the data frames.
`. Therefore, each data frame has as many slots as there are
`stations
`
`Each earth station lkeeps a reservatiOn table to track the
`data slot allocation. To make reservation for data slots, the
`earth station transmits its “new reservation” count in its
`reservation slot. The stations which do not have data to send
`place a value of zero in their fixed reservation slots. The new
`reservation count represents the number of data packets that
`arrived after the last reservation took place. All the earth
`stations receive the reservation packet and adjust
`their
`reservation table values by adding the new reservation counts
`at a globally agreed upon time.
`The allocation of data slots now becomes straightforward.
`Those stations whose reservation table entries are not zero
`transmit their data packet in their fixed slots. The dataslots
`which belong to station with no data packets are assigned in a
`round-robin manner among those stations with outstanding
`data packets. The sender for each slot is determined just prior
`to the slot transmission time. In this scheme, synchronization
`is acquired and maintained by having each station send its
`own reservation table entry in its reservation slot.
`
`CONFLICT-FREE MlULTIACCESS (CFMA)
`This scheme [6] eliminates all conflict on the satellite
`multiaccess channel. The channel is divided into frames.
`Each frame is subdivided into an R-vector, anA-vector, and
`an I-vector. Fig. 4 shows the frame structure and three
`vectors. The R-vector is used to request future reservations
`and is divided into a number of reservation slots. The number
`of reservation slots in the R-vector is equal to the number of
`earth stations. Each one of the earth stations is assigned a
`reservation slot in the RT-frame. This avoids contention for the
`reservation slot. The /\-vector is divided into a number of
`mini-slots which are . used to send acknowledgment
`for
`previously received packets. An I-vector in a frame is divided
`into data slots. In this scheme, the maximum number of slots
`a station may request is equal to the number of slots in the I-
`
`
`Lasageeggaw
`FE,M.M.-WW;WWWi
`
`
`‘
`3
`J6—
`:‘RESERVATON store ’
`
`sew—m- RE§5RWIQNFRAME —-—-—«———-ef
`Fig. 3. R-TDMA channel.
`
`19
`
`
`
`
`
`Flg. 4. Frame structure tor CFMA channel.
`
`frame. Assuming that there are m data slots in an l-frame, the
`allocation of data slots is based on assigning a priority order
`for each of the m slots. For example, if the number of stations
`equals the number of data slots (N = m) in an I-vector, then
`the priority order for each data slot is different. Every station
`has one data slot for which it has first priority, another for
`which it has second priority, and so on down to the least
`priority. If a station does not use its first priority data slot, then
`a station with second priority to that slot gets a chance to use
`that data slot. If all stations are busy, then each of the stations
`will be allocated its first priority data slot and no station will be
`allocated more than one slot in the above example. The
`overhead involved in this system does not seem to be high in
`terms of channel bandwidth.
`
`CONTENTION-BASED DEMAND '
`ASSIGNMENT PROTOCOL (CPODA)
`This protocol is designed to handle packetized data and
`voice traffic [7]. It can handle traffic with multiple priority and
`delay class distinctions, variable message lengths, and
`arbitrary load distribution among the stations.
`The channel is divided into fiifed size frames, and each'
`frame consists of reservation and information subframes. The
`reservation subframe is divided into fixed size reservation
`
`slots. in this scheme, the reservation subframe is allowed to
`grow or shrink according to the amount of traffic. Therefore,
`when the number of reservations for the information frame is
`zero, the reservation subframe expands to occupy the whole
`frame. On the other hand, when the system is fully loaded, the
`reservation subframe contracts to the‘minimum number of
`slots required to allow reservations by high priority traffic or
`previously idle stations.
`There are two ways in which reservation for information
`subframes can be done. The first way is to send a reservation
`in the slots in the reservation subframe. The stations use
`
`contention to gain access to the reservation slot. The sec'ond
`way is to send the reservation by piggybacking them in the
`
`4
`
`
`
`
`IEEE COMMUNICATIONS MAGAZINE
`
`the reserved message transmission. A
`field of
`header
`maximum‘of only two new reservations is allowed in each one
`of the messages. This allows a station transmitting messages
`to use the piggybacking technique to build their reservation,
`thereby leaving the reservation subframe free for new entries
`and/or higher priority traffic.
`A distributed control is used to schedule channel time for
`each earth station to transmit messages. The scheduling is
`done by forming a queue of the desired transmissions from the
`explicit reservation requested by the stations. The channel
`scheduling in this scheme is some function of message priority
`and delay. Thus, a low priority message with a short delay
`constraint may typically be serviced before a high priority
`message with a long delay. The ordering, to some extent, is a
`weighted function of priority and delay.
`Each station carries out a consistency check to assure
`scheduling synchronization. A station is in synchronization
`when its scheduling decision agrees with the actual transmis-
`sion in the channel. A station can be in one of three states as
`shown below.
`i
`
`ACQUISITION
`
`INITIAL
`
`
`
`channel. If the station, in the monitoring channel, finds itself in
`synchronism again within a fixed period of time, it moves
`back to the in-isync state and participates in message
`transmission. Otherwise,
`it moves to the initial acquisition
`state. In this state, the statiOn listens to the new reservations
`on the channel and builds up its channel scheduling
`information. The station does not transmit any message.
`Once this station has constructed a reservation list compatible
`with other stations, it can move to the out-of—sync state. 1
`
`CONCLUSIONS
`
`A number of multiple access protocols have been
`presented, some of which are undergoing testing for satellite
`communication. These reservation methods provide-ameans
`to increase channel utilization compared to nonreservation
`schemes. In all the schemes, one must trade off complexity of
`implementation with suitable performance. Therefore,
`in
`the final analysis, it is cost which will dictate which of the
`protocol schemes is suitable for a particular application.
`REFERENCES
`
`‘
`
`[1] N. Abramson, “Packet switching with satellites," in Proc. AFIPS
`Conf., vol. 42, June 1973.
`[2] R. Binder, “A dynamic packetvswitching system for satellite broadcast
`channel,” in Proc. ICC '75, San Francisco, CA, June 1975.
`[3] W. R. Crowther et al., “A system for broadcast communication:
`Reservation-ALOHA," in Proc. 6th Hawaii Int. Conf. Syst. Sci.,
`Jan. 1973.
`[4] D. W. Davis et al., Computer Networks and Their Protocols. New
`York: Wiley, 1979.
`[5] M. Gerla and L. Kleinrock, “Closed loop stability control for 5»
`ALOHA satellite communication," presented at' the 5th Data
`Commun. Symp., Sept. 1977,
`[6] H. R. Hwa, “A framed ALOHA system," in Proc. PACNETSymp.,
`Sendai, Japan, Aug. 1975.
`[7] 1. Jacobs 2! 'ol., “CPODA—A demand asssignment protocol for
`SATNET," in Proc. 51h Data Commun. Symp., 1977.
`[8] l. M. Jacobs el al., “General purpose packet satellite network,"
`Proc. IEEE’NOV. 1978.
`[9] L. Kleinrock and S. S. Lam, “Packet—switching in a slotted satellite
`channel,” in Proc. AFIPS Cont, vol. 42, June 1973.
`[10] S. Lam and L. Kleinrock, “Packet switching in a multi-access broad-
`cast channel: Dynamic control procedures," IEEE Trans. Commun.,
`vol. COM-23, Sept. 1975.
`[11] L. G. Roberts. “Dynamic allocation of satellite capacity through
`packet reservation," in Proc. AFIPS Cont; vol. 42, June 1973.
`[12] ”Satellite carrier posed for increasing demands,” Commun. News,
`Mar. 1979.
`‘
`[13] F. A. Tobagi et al., ”Modeling and measurement techniques in packet
`communication networks,” Proc. IEEE, vol. 66, Nov. 1978.
`[14] R. Weissler et al., “Synchronization and multiple access protocols in
`the initial satellite IMP,” in Proc. COMPCON, Fall 1978.
`
`
`I
`
`A station in the in-sync state is in synchronism with the
`actual transmission taking place in the channel. Hence, it can
`continue sending messages at the scheduled time. Whenever
`a station detects a number of inconsistent scheduling within a
`specified time period, it moves to the out-of-sync state. In this
`state, the station is not allowed to send any message; instead it
`carries out channel scheduling and closely monitors the
`
`for Computing Machinery and the IEEE.
`
`C. Retnadhas received the Ph.D. degree from
`Iowa State University. He is a faculty member
`in computer science at Western Illinois Univer-
`sity. His research interests include computer
`architecture, computer communication net-
`works, and distributed processing.
`Dr. Retnadhas is a member of the Association
`
`
`
`20
`
`
`5
`
`