`
`141
`
`Data packet
`
`Req uest packet
`
`undifiodeni Req I!) .
`EPream- Req Info CRC
`
`Data Pklfor
`req
`in old l'r
`
`recently proposed alternative approach [37] is described which does not require
`
`scenario uses slotted ALOHA for access [34—36] (as shown in Fig. 7.12), so that
`the reservation overhead required is independent of the actual number of
`terminals supported. Although Fig. 7.12 shows a system based on a TDMA-like
`frame, an interleaved format may also be used to minimise framing latency
`efi'ects. Note that, in principle, it is also possible to use other slotted contention
`mechanisms (such as the tree CRA) for the reservation messages. It is observed
`that, since contention access is characterised by a relatively low capacity, the
`allocation of channel time for reservations could become significant unless the
`ratio of request message to data message length is quite large.
`Typical achievable capacities for VSAT traffic parameters are in the range of
`04- to 0-6, depending on the length distribution of message traffic to be supported.
`Slotted ALOHA access generally leads to more effective coverage of a range of
`VSAT traffic profiles than the TDMA access case discussed above. As for all
`DAMA protocols, this protocol is characterised by a high minimum delay of
`~ 0-6 s, ofi‘set to some extent by low delay variance. As for the TDMA access case
`discussed above, the throughput and delay variance advantages ofDAMA are at
`the expense of higher implementation complexity and poorer robustness. Never-
`theless, since DAMA with slotted ALOHA access provides good overall perforu
`mancc and can handle mixed interactiveffile-transfer tral’fic, it is an important
`Candidate for many VSAT applications.
`
`propagation
`deiay (+1 frame)
`Rel: delay for R1“)
`Reu delay for RZU)
`
`Channel
`propagation
`delay [‘1 frame]
`Rel: delay for R311)
`Rel: delay for 111(2)
`
`Legenn:
`C] Request
`. Collision
`
`D Data packet
`
`Fig. 7.12
`
`Illustration of DAMA with slotted ALOHA arrest
`
`7.2.3.3 Unslotted locally synchronous reservation Traditionally, controlled access has
`been associated with time slotted channels because of the need to identify and
`allocate channel time segments to stations on demand. In this subsection, a
`
`39
`
`
`
`142 Mulliaccesr protaculsfor VSAT networks
`Asyncn Access Hodemm) ..LSRH..ARH LSRH .
`..ARn.
`,
`Res erv
`Reserv
`
`.
`
`,,
`
`,
`
`Locally Synchronous Reserved
`node (LSRH)
`
`RP~B 8‘ RP-C
`,
`retransmitted in ARM
`wltn random delay
`
`LSRM
`ends
`
`Scheduled LSRH for
`Reserv Pkts A & D
`
`Channel propagation delay-F
`
`/
`
`TRANSHIT
`time
`
`
`
`RECEIVEtime
`
`reservation schemes.
`
`TDMA—like timing, and is classified as ‘locally synchronous’ or ‘self—synchronis—
`ing’. Protocols of this type were conceived as semi—compatible upgrades ofunslot—
`ted random access protocols such as ALOHA, SREJ—ALOHA or time-of-arrival
`CRA, for use in environments with a high proportion of long messages or mixed
`interactive/file-transfer traffic.
`The principle in locally synchronous reservation systems is to provide initial
`access for request packets in an unslotted mode such as ALOHA. When a
`specified number of successful requests (K > 1) have been received, the channel
`switches to the locally synchronous reserved message transmission mode. As in the
`time-of—arrival ORA, scheduled transmissions are locally synchronised using
`channel event based timing. An example of the operation of such a protocol using
`ALOHA access is shown in Fig. 7.13. For typical ratios of data packet
`to
`reservation packet length, achievable capacity is of the order of Geo-7 with
`ALOHA access and 0-8—0-9 with time—of—arrival CRA access (applicable to fixed
`length packet formats only). Since partitioning of channel time is dynamically
`achieved in these protocols, they tend to have superior delay-throughput charac-
`teristics when compared with conventional TDMA based systems. Of course, the
`irreducible reservation latency delay is also a characteristic of these protocols, as
`indicated in Fig. 7.2.
`
`Fig. 7.13
`
`Illustration of channel events in locally synchronous reservation with ALOHA
`access
`
`D Data packet
`[:l Reservation pk:
`Collision
`Locally synch
`timing marker
`
`A
`
`7.2.3.4 Hybrid reservation/random access The demand assignment approaches des—
`cribed above are suitable for traffic scenarios with a relatively high proportion of
`long messages. However, for traffic profiles with a mix of short interactive and
`long transaction/batch messages,
`the use of pure DAMA results is obviously
`inefficient for the short messages, and is associated with a high latency delay. On
`the other hand, random access systems, which are well suited to interactive traffic,
`are generally quite inefficient for mixed traflic environments. These considera—
`tions motivate hybrid reservation/random access schemes which combine the low
`delay advantages of random access with the high throughput properties of
`
`
`
`40
`
`
`
`Multiaccem protocolsfir VSA T networks
`
`143
`
`Several approaches have been proposed for this purpose [38411]. The simplest
`strategy is to transmit short messages in the reservation slots directly, without
`reservation delays, and to transmit a reservation message (which could be ‘piggy-
`backed’ with short data packets) only if additional capacity is required. One
`could also establish procedures in which there is no explicit reservation of data
`slots, but where successful random access in a frame results in an implicit reserva—
`tion of the same slot in future frames. These protocols offer the possibility of high
`DAMA throughput ( > 06) along with low access delay for interactive messages,
`potentially leading to throughput—delay characteristics which transition grace-
`fully f‘rom the low-dflow-S random access curves to the high-djhigh-S reservation
`curves in Fig. 7.2. However, optimisation of performance (while avoiding subtle
`instability and deadlock conditions) over a range of traffic profiles may be
`difiieult to achieve in practice.
`
`7.3 Performance comparison of candidate VSAT
`protocols
`
`In this section, a. more detailed quantitative performance comparison of can—
`didate VSAT multiaccess protocols is presented, chosen from those discussed in
`Section 7.2. Specifically, based on a variety of implementation and performance
`factors, unslotted ALOHA, SREjiALOHA, slotted ALOHA and DANIA with
`slotted ALOHA access have been chosen for the detailed comparison. Note that
`all the techniques considered in this section are ‘rnature= and have been validated
`by a variety ofindependent analytical and simulation studies. Although analyti-
`cal tools for performance evaluation are available for each ofthe above four access
`methods, direct event simulation is used for the results presented here. This makes
`it possible to obtain delay distributions, which are important for data network
`design, as well as to incorporate realistic traffic models which must generally be
`approximated in the analytical models. Detailed discussion of the simulation
`models used can be found in Reference 42.
`
`7.3.1 VSAT traffic models
`
`,
`
`Remote stations which support interactive data applications typically generate
`short, variable length messages with average data rate orders ofmagnitude lower
`than the multiaccess (terminal to hub) channel speed. Roughly speaking, the
`remote station trafiic model is described by two attributes: (i) the average rate at
`which new messages are generated, and (ii) parameters specifying the length
`distribution. Note that (i) and (ii) together can be combined to determine the
`average data rate per station in bits per second. In Fig. 7.14, approximate regions
`in the average data rate versus average message length plane that are occupied
`by some common VSAT applications are. shown. It is observed that the impor-
`tant interactive data scenario is in the region near the origin, corresponding to
`low average data rate and short messages. Note that although, in Fig. 7.14, the
`length distribution has been approximated by a single parameter which is the
`average value, in general, an accurate evaluation requires exact determination of
`message length distribution for a particular application. It is often reasonable to
`
`Section 7.3 and Figs. 714—20 arefrom: MTGHAUDHURI, D., and jOSEPH, K: ‘C’annnet’
`access protocols for VSAT networks:
`a
`comparative
`soalualian’,
`IEEE Communications
`hiagazine, Special series on VSA '1", Ala} 1988, pp. 34744
`Reproduced by permission of IEEE ((7) IEEE 1988
`
`‘I
`l
`‘lrl-.Il
`
`41
`
`
`
`_. %
`
`‘2;
`
`.. 0
`
`%0!56kbills
`
`Averageterminalrale,bits/s
`
`Fig. 7.14
`
`Message length (chars)
`Tmfiic source parameter regionsfar potential VSA T applications
`((3 IEEE 1988
`
`each station.
`The parameters used in this performance comparison correspond to a typical
`interactive transaction application from the parameter region identified in Fig.
`7.14, and are summarised for convenient reference in Table 7.3.
`
`I44- Multiaccesr protocolsfor VSATnetworks
`10‘
`
`
`
`
`
`7.3.2 C/zarmel and protocol parameters
`
`many
`detailed nature of the arrival
`that need to be considered in
`'
`process (which is, in gener
`the evaluation. Typically, each remote station is connected to a cluster controller
`that supports several interactive data terminals, which do not contribute traffic
`to the VSAT while they are waiting for a reply from the host computer. Detailed
`models for the general case, with M 2 1 terminals
`the VSAT, have been considered [4-3]. For sim
`
`42
`
`
`
`=0
`
`%of55kbit/s
`
`i,
`
`-e and
`-=unica-
`... ~| IiStiC
`'
`'
`the
`many
`_-aJTival
`‘
`.. ’I in
`'
`‘r ller
`
`Summary qf channel parameters wed in performance comparison
`Table 7.4
`© IEEE 1988
`
`Channel data rate
`
`Channel symbol rate (rate 172 coded)
`Minimum lE,,lN0 required
`Modern acquisition preamble
`One way propagation delay
`Minimum ACK delay (2 hops)
`
`= 56 kbit/s
`
`=112kbit/s
`=8-O (38
`=32 bits (4 characters)
`: 0-2? 3
`
`Multiaccess protocols for VSA T networks
`
`Summay; of trqfiic source parameters used in performance
`Table 7.3
`comparison CC) IEEE 1988
`
`= 1
`=250msglh = 0-07 meg/s
`
`interactive terminals per VSAT
`New message rate per terminal
`(unblocked mode)
`New message rate per terminal
`(blocked mode)
`—truncated exponential
`New message length distribution
`2100 chars (800 bits)
`Average new message length
`=256 chars (204B bits)
`Maximum new message length
`
`
`protocols under consideration have been selected according to established design
`principles, and are given in Table 7.5. Note that, as discussed earlier, the key
`retransmission delay parameter is chosen for minimum stable delay, and is
`optimised at each level of load, N. For ALOHA, the retransmission delay is the
`only significant parameter of the protocol. However, for slotted ALOHA and
`SREJ—ALOHA, the choice ofappropriate packet sizes is also important. For slotted
`ALOHA, the convenient approach is adopted of selecting a message packet size
`(Le. slot size less guard time and all overheads) equal to the maximum possible
`message length. For the present example, referring to the traflic model in Table
`7.3, this is assumed equal to 256 characters, as might be the case in many practical
`systems. For SREj—ALOHA, subpacket size is an extremely important choice: in
`general, optimum subpacket size depends upon the acquisition and addressing
`overhead, but is also a load dependent quantity. Fortunately, it can be shown
`that, for the present set of parameters, there is a fairly robust choice of subpacket
`size (50 characters for an average message length of 100 characters, as given in
`Table 7.5), which is close to optimum at moderate to heavy channel load. For
`DAMA/TDMA, several parameters such as the fraction of time allocated to
`reservation traflic,
`the average retransmission delay for colliding reservation
`messages and the message and reservation slot sizes need to be specified. The
`design procedure used here is based on optimally allocating a fraction of channel
`capacity to the reservation channel, noting that total delay is the sum ofreserva-
`tion access delay and message assignment delay. As discussed above for the
`ALOHA protocols, the retransmission delay for the slotted ALOHA reservation
`subchannel is selected for minimum stable delay. As for the slotted ALOHA case,
`the DAMA/TDMA system is based on a message slot size equal to the maximum
`
`=0‘54S
`
`43
`
`
`
`
`
`
`ALOHA
`
`146 Maritime?” protocolrfar VSAT networks
`Table 7.5
`Summary ofprotocolparameters wed in pejormance comparison © IEEE 1988
`General
`Link level overhead per packet or subpaoket (L2) = Tchars
`Network level overhead per message (L3)
`= 8chars
`Minimum ALOHA mode retransmission delay
`= 0-675
`
`Retransmission delay optimised for deiay at each traffic load, under constraint of stable
`
`operation.
`
`
`Single variable length packet per message
`ALOHA packet: data length + L2 + L3 + preamble
`’_’______________———————
`SHEJ—ALOHA
`
`Multiple fixed length subpackets per message
`Subpacket size optimised for each average message length. L
`SREJ subpacket: fixed data length + L2 + preamble
`Subpacket size: S = 40 char for L = SOChar
`50
`100
`60
`150
`
`data + L2 + L3 + preamble
`
`
`
`
`
`Slotted ALOHA
`Single fixed length packet per message
`Slotted ALOHA packet: data length + filler
`E 275 chars (fixed)
`Time slot size: packet length + guard time (4chars) = 279 chars
`___________—_——————’———
`DAMA! TDMA (slotted ALOHA reservation)
`Reservation packet length 26 chars (including preamble)
`Reservation slot length: reservation packet length + guard time = 30 chars
`DAMA message packet: data length + tiller data + L2 + L3 + preamble
`= 275 chars (fixed)
`DAMA message slot size: message packet length + gu
`Assignment delay = 0-673
`Fractional allocation of message and reservation slots optimised f
`Message and reservation slots interleaved
`
`ard time (4 chars) : 279 chars
`
`or minimum delay.
`
`
`
`
`
`actcrs. In general, DAMA efficiency can be improved
`
`message length of 256 char
`slots, but this is at the expense of some increase in
`by using shorter message
`implementation complexity.
`
`7.3.3 Numerical results
`
`Fig. 7.15 shows the familiar throughput—(:lela}r characteristics ofthe four protocols
`under consideration. Note that throughput shown is the useful data throughput,
`
`after accounting for overheads such as acquisition preamble in ALOHA, guard
`(1 slot time (due to variable message length) in
`time} acquisition time and waste
`slotted ALOHA, subpacket overhead in SREj—ALOHA and reservation channel
`
`44
`
`44
`
`
`
`0‘15
`
`Throughput
`
`Fig. 7.15 Daley-throughput characteristicsfor candidate multimms protocols © IEEE 1988
`
`Multiaccerr protocolsfor VSAT networks
`
`147
`
`SREJ-ALOHA
`
`becomes overloaded. In Table 7.6, a comparison is presented of protocol capaci—
`
`overhead and wasted message slot time in DAMA/TDMA. The delay values
`include the effect of slot synchronisation latency delays for the slotted protocols.
`It is observed from the figure, that, for this message model, slotted ALOHA
`performs only slightly better than ALOHA, while both are considerably outper-
`formed by SREJMALOHA. Obviously, the relative performance is strongly de-
`pendent on message length, an effect which should be investigated at a later
`point. Fig. 7.15 also shows that the maximum throughput achieved by the
`DAMA approach is higher than that of the random access protocols, but is
`achieved with a high irreducible delay of about 0-75 5.
`As discussed earlier, a more useful characterisation, as in Fig. 7.16, is provided
`by plotting curves of average delay and peak (95th percentile) delay as a function
`of the number of VSAT stations supported on the same channel, N. Since
`interactive network design specifications must include average as well as peak
`delay constraints, such a set of curves is a prerequisite for determining the
`“capacity”, i.e. the number of remote stations (VSATS) supported on a channel
`while satisfying the performance requirements, JV“. The operating value of N*
`is determined by the tighter of the average and peak delay constraints. From Fig.
`7.16, it can be observed that peak delay for ALOHA tends to rise rather rapidly,
`while DARIA is characterised by high average delay and low delay variance.
`Slotted ALOHA and SREJ—ALOHA provide better peak delay properties than
`ALOHA, but are also characterised by rapidly rising peak delay as the channel
`
`45
`
`
`
`148 Multiaccesr protooobfor VSAT networks
`10
`
`Slotted ALOHA
`SREJ-
`ALOHA
`
`95 percentile
`
`- - -- - Average
`
`Number of VSATS
`
`Fig. 7.16 Average and peak delay) versus number of VSA T3 for candidate multioccess
`protocols. © IEEE 1988
`
`ties with delay constraints of (0-755 average, 2-55 peak) and (1 s average-x33
`peak). Observe that the 0-75 s average delay case is not achievable with DAMA,
`but when the average delay requirement is relaxed to l s, DAMA supports many
`more terminals than the contention access techniques. Comparing the ALOHA
`
`Table 7.6 Channel capacities for candidate protocols under alternative delay comtmz'nts
`© IEEE 1988
`
`Multlaccess
`protocol
`
`Channel capacity (number of VSATS supported. N")
`
`Delay constraints:
`0-753 average
`2-55 peak (95%)
`
`Delay constraints:
`103 average
`3-05 peak (95%)
`
`
`
`80
`75
`ALOHA
`110
`105
`Slotted ALOHA
`140
`130
`SREJ—ALOHA
`230
`*-
`DAMAITDMA
`—__——-—_
`* Objective not achievable
`
`
`
`Hill"
`
`l
`
`i
`
`46
`
`46
`
`
`
`
`
`
`
`Number of remote stations
`
`1 50
`
`Fig. 7.17 Variation of total trqfic with number ty’ VSA Ts fir the diferent protocols
`((3 IEEE 1988
`
`Multiaccesr pmtocolsfar VSAT networks
`L = 100
`
`14-9
`
`,’ DAMA
`, (S-ALOHA
`
`I
`
`Totalchanneltraffic
`
`the delay penality experienced by long messages can be rather significant, par-
`
`protocols, it is observed from the table that, at both performance levels, SREJ—
`ALOHA supports about 25% and 75% more stations than slotted ALOHA and
`ALOHA, respectively.
`Curves of total channel traffic (G) versus the number of remote stations per
`channel (N) are given in Fig. 7.17. These curves are ofsome ancillary importance
`because they relate to the ‘power efficiency” ofa protocol, which needs to be
`considered in power limited situations that frequently characterise VSAT
`networks. From the figures, observe that, at any value of JV, ALOHA has the
`lowest total channel traflic (and hence the best power efliciency), followed by
`SREJ-ALOHA, DAMA and slotted ALOHA, in that order. Note also that
`random access systems display a rapid increase in G after reaching their respective
`congestion points, clue to a non-linear increase in retransmission traffic.
`Further characterisation of the protocols in terms of delay distributions at
`specified load levels is given in Fig. 7.18. In the figure, delay distributions are
`plotted for ALOHA, slotted ALOHA, SREJ—ALOHA and DAMA, designed to
`operate at an .N* corresponding to an average delay of l s (as determined from
`curves in Fig. 7.16). Observe that, as expected, DAMA demonstrates relatively
`low delay variance, while ALOHA is characterised by very high delay variance,
`mainly due to repeated collisions experienced by long messages. Slotted ALOHA
`and SREJ—ALOHA display moderate delay variance, although substantially
`greater than that of DAMA. SREj—ALOHA appears to have lower ‘tails’ of the
`delay distribution than the other contention access techniques. A second detailed
`characterisation of interest is the plot of average delay versus message length
`shown in Fig. 7.19. Observe that DAMA and slotted ALOHA, for which the slot
`size has been chosen to equal the maximum message length, have delays which
`are independent of the actual length. In contrast, ALOHA and SREj—ALOHA
`accommodate variable length transmissions, so that longer messages tend to have
`a higher collision probability (and hence delay) than short messages. In ALOHA,
`
`47
`
`
`
`.
`
`A.
`E”>.
`
`E 3Ea
`
`)>
`<
`
`Fig. 7 L..,
`
`consil -
`accefi -.
`ALOHA
`more ..
`ALO
`advam»
`can be
`has also
`
`ately '
`
`-
`
`Table
`with m .
`
`Message
`(ch.
`.
`
`
`
`150 Multz'accem protamlsfar VSA T networks
`
`
`
`CDF
`
`Mean delay = 1s
`
`SHEJ-ALOHA
`
`0
`
`2
`
`4
`
`S
`Delay (5)
`
`3
`
`10
`
`12
`
`Fig. 7.18 Deity; distributions for candidate preface]; at channel load correspana'ing to an
`average delay ofI-Os. © IEEE 1988
`
`ticularly at high channel load. However, the SREJ mechanism in selective reject
`ALOHA overcomes this problem to a large extent, so that the curves Show only
`a modest increase in delay with message length.
`The above results in Figs. 7.15—7‘19 were based on a specific VSAT traflic
`source model, with an average of 250 messages/hour and truncated exponential
`messages with maximum 256 characters and average 100 characters. In order to
`study the sensitivity of protocol performance with respect to average message
`length, Table 7.7 is introduced which shows the number of VSATs supported per
`channel (based on a 105 average delay criterion), for average message length
`equal to 50, 100 and 150 characters respectively (truncated exponential distribu-
`tion assumed). The results show that, as expected, slotted ALOHA and DAMA
`have capacities which are insensitive to the message length, while both ALOHA
`and SREJ—ALOHA exhibit decreasing capacity as message length increases. In
`general, it is found that SREJeALOHA has the highest capacity for short to
`medium length, while for message lengths of 150 characters or higher, slotted
`ALOHA is the preferred protocol. Overall, Figs. 721577.19 provide a detailed
`comparison of the candidate VSAT protocols considered. In general, contention
`techniques are characterised by curves of delay versus number of VSATs which
`start at the origin (Le. low delay at low throughput), whereas DAMA systems
`exhibit significant latency (irreducible) delay. This may indicate the need to use
`random access for applications requiring very low access delay, even before
`
`48
`
`
`
`Multiaccass firotacolsfir VSAT networks
`
`151
`
`—— 125 Remote stations
`
`_ _ — —-
`
`75 Remote stations
`
`
`
`Averagedelay(a)
`
`1 i
`
`\
`
`Slotted ALOHA
`
`_ _ _ _ _ _ _ _ EMA—(1.09 RimPte—S‘E‘ifflsl
`SREJ-ALOHA
`
`_
`
`__.___._::._____,._—=_.__._._=__
`
`‘_ \ SREJ-ALDHA
`
`
`
`0-01
`
`0-02
`Length (5)
`
`-
`
`0-04
`
`Fig. 7.19 Variation of average delay with message length for the candidate protocols
`© IEEE 1988
`
`Table 7.7 Variation of channel capacity (with average delay constraint equal to 1:05)
`with average message length. © IEEE 1988
`5—m—
`
`lllllllllllllllllll
`
`considering implementation complexity factors. When selecting between random
`access protocols, it has been shown that, for practical parameter ranges, SREJ—
`ALOHA is superior to both ALOHA and slotted ALOHA (which is definitely
`more complex in terms of implementation). For short message environments,
`ALOHA may also be used, especially because it offers significant implementation
`advantages. Slotted ALOHA is generally outperformed by SREj—ALOHA, but
`can be used in environments with a large fraction of long fixed size messages. It
`has also been observed that SREjfiALOHA and slotted ALOHA offer approxim-
`ately equivalent delay distribution efl'eets, while ALOHA is generally charac-
`
`Message length
`(characters)
`
`Channel capacity {number of VSATS supported, N")
`
`DAMA/
`SFlEJ—
`Slotted
`TDMA
`ALOHA
`ALOHA
`———_—‘——*_
`
`ALOHA
`
`50
`
`250
`205
`130
`250
`160
`130
`250
`115
`130
`75
`am—
`
`49
`
`
`
`152 Multiaccesr protocols for VSAT networks
`
`
`the ‘peak’ delay
`terised by long delay tails. Depending on the application,
`
`performance such as those shown in Fig. 7.16 is also an important consideration
`
`for network design. In applications requiring very low peak delays, ALOHA may
`
`not be a suitable protocol, and either SREJ—ALOHA or slotted ALOHA should
`
`be used. For scenarios in which relatively high ( > 0-75 5) average access delay is
`
`permissible, DAMA with slotted ALOHA access may provide capacity advan-
`
`tages. In addition, suitably designed DAMA protocols tend to provide relatively
`
`low ratios of peak delay to average delay, which may be an advantage in some
`
`applications. However, the implementation complexity, which involves slotting,
`
`framing, service queue management and control signalling,
`is considerably
`
`greater than for any of the contention access techniques. Thus, the requirement
`
`for low cost VSAT equipment is likely to motivate the use of contention access
`
`for first generation VSAT systems, except in traffic scenarios for which the
`
`capacity penalty would be extremely severe. Moreover, as discussed in the earlier
`
`section, there are various approaches to semi-compatible upgrades of contention
`
`protocol equipment for future trafiic scenarios which require a reservation mode.
`
`To complete the discussion, an attempt will be made to provide a qualitative
`
`comparison between multiaccess protocols over the range of traflcic source para-
`
`meters shown in the terminal—rate/message~length diagram shown in Fig. 7.14.
`
`The results presented so far apply to the interactive (transaction or point—of—sale)
`
`environment shown near the origin of Fig. 7.14. Widening the scope of com-
`
`parison to the entire region shown in the Figure, it would be useful to specify
`
`optimum regions for the various alternative VSAT protocols. Accordingly, Fig.
`
`7.20 provides a qualitative assessment of optimum regions for the VSAT protocols
`
`discussed. An attempt has been made to incorporate the issue of satellite power
`
`as well as bandwidth, without entering into the details, which are discussed in
`
`Chapter 9. Specifically, Fig. 7.20;; shows the optimum protocol regions“ for a
`
`moderately power limited example, as might be encountered in a system with
`
`small antenna size (1-2 m), high ratio of hub outbound traffic to inbound traffic
`
`and/or high availability requirements. Fig. 7.205 shows a similar plot for a
`
`moderately bandwidth limited system, which may occur in a situation with larger
`
`antennas (1-8 m or greater), powerful forward error corrected modems, balanced
`
`outbound and inbound traflic and/or low availability requirements.
`
`Observe from the two Figures that, in general, ALOHA-type protocols are
`
`optimum for scenarios with short messages and low average data rate per remote
`
`station, in agreement with conventional wisdom. DAMA is recommended for
`
`higher message lengths, but for low to medium station traflic volume, while
`
`TDMA is the choice for systems with high station traffic volume, independent of
`
`message length. Note that there is considerable overlap between optimum regions
`
`for the power limited scenario since bandwidth use is relatively non—critical. On
`
`the other hand, for the bandwidth limited case, there is little overlap between the
`
`optimum regions for each of the protocols, and this can be attributed to the
`
`greater importance of the multiacccss efficiency in determining system capacity.
`
`
`
`'Note also that TDMA, which is a useful protocol for medium to high volume stations, is
`now included. Also the three ALOHA protocols have been lumped together, with the
`understanding that the choice between these will depend upon the message length regime
`
`of interest along with implementation factors.
`
`
`
`w
`
`W
`
`x
`
`50
`
`50
`
`
`
`.923.2!35:5mums;
`
`$.25.99E5852mamhwni
`
`Message length (chars)
`
`Fig. 7.20 Optimum protocol regiomfnr typical Ku-band satellite scenarios © IEEE 1933
`a moderately power limited case
`b moderately bandwidth limited case
`
`7.4 Conclusions
`
`A review of established and new approaches to satellite multiaceess for VSAT
`lications has been presented. A variety‘of contention and reservation based
`has been described and
`protocols for use on both slotted and unslotted channels
`
`ma.da1.”m
`
`flmCmY.
`
`.0.
`
`ksTowz8RTmVv!
`
`153
`
`ALOH
`
`Manxmm.3axe
`
`aPP
`
`\m\l
`
`51
`
`
`
`154- Multiamess protocolsfor VSA T networks
`
`compared in terms of key attributes such as throughput, delay, stability, robust-
`ness, operational convenience and implementation complexity. After the survey,
`detailed performance results for four candidate ‘first generation’ VSAT protocols
`were presented (ALOHA, selective reject ALOHA, slotted ALOHA and DAMA
`with slotted ALOHA access) applied to an example transaction application. It
`has been shown that, among the random class systems considered, SREJ—
`ALOHA generally outperforms both ALOHA and slotted ALOHA. DAMA is
`shown to achieve a higher capacity and lower delay variance than the random
`access alternatives, but this is at the expense of at 0-755 irreducible delay, poor
`robustness and higher implementation complexity. In view of the relatively low
`impact of VSAT to hub space segment cost on overall system economics, it is
`expected that delay, implementation complexity and robustness will be primary
`considerations in VSAT access protocol selection, with capacity (within reason-
`able ranges) being an important, but secondary issue.
`
`H"
`
`1 METCALFE, R.M., and BOGGS, D.R.: ‘Ethernet: Distributed packet switching for
`local computer networks’, Communications ACM, July 1976, pp. 395—4114
`BUX, W., at all: ‘A reliable token system for local area communication’, National
`Teiemmmunimtians Conference, Dec. 1981, pp. 142.21%
`ABRAMSON, N; ‘The ALOHA systemiAnother alternative for computer com-
`munications’. AFIPS Conf. Prat., 1970, 37, pp. 281—285
`TOBAGI, F.A.:
`‘Multiaccess protocols in packet communication systems’, IEEE
`Trans. Commun, April 1980, pp. 458488
`LAM, 8.8.: ‘Satellite packet communication— Multiple access protocols and perfor-
`mance’, IEEE Trans. Commun, Oct. 1979, pp. 1456—1466
`SCHWARTZ, j.W., AEIN, J.M., and KAISER, J; ‘Modulation techniques for
`multiple access to a hard limiting repeater’, Pros. IEEE, May 1966, pp. 763—777
`PURSLEY, M.B., and SARWATE, D.V.: ‘Performance evaluation for phase-coded
`spread spectrum multiple—access communicatioanart 1: System analysis’, IEEE
`Trans. Commun., Aug. 1977, pp. 795e799
`. GERANIOTIS, E.A., and PURSLEY, M.B.: ‘Error probabilities for slow-frequency-
`hopped spread-spectrum multiple-access communications over fading channels’, IEEE
`Tram. Commem, May 1982, pp. 99671009
`9 SCHMIDT, W.G.: ‘Satellite time division multiple access systems: Past, present and
`future’, Telecommun, 1974, 7, pp. 21724
`10 LAM, S.S.: ‘Delay analysis of time division multiple access (TDMA) channel’, IEEE
`Trans. Commun, Dec. 1977, pp. 1489-71494
`1 l KOBAYASHI, H., ONOZATO, Y., and HUYNH, D.: ‘An approximate method for the
`design and analysis of an ALOHA system’, IEEE Tram. Commun.,]an. 1977, pp. l49r157
`12 JENQ, Y.C.: ‘On the stability ofALOHA systems’, IEEE Trans. Commun., Nov. 1980,
`pp. 1936—1939
`13 BELLINI, S., and BORGONOVO, F.: ‘On the throughput of an ALOHA channel
`with variable length packets’, IEEE Tram. C0mmun., Nov. 1980, pp. 1932—1935
`14- RAYCHAUDHURI, D.: ‘ALOHA with multipacket messages and ARQ—type retrans-
`mission protocols—Throughput analysis’, IEEE Trans. Commun, Feb. 1984, pp. 1487154-
`15 RAYCHAUDHURI, D.: ‘Stability, throughput and delay of asynchronous selective
`reject ALOHA’, IEEE Tram. Cammun., July 1987, pp. 767’ .772
`16 MASSEY, J.L., and MATHYS, P: ‘The collision channel without feedback’, IEEE
`Tram. Inf. 77290131, March 1985, pp. 192—204-
`17 ROBERTS, L.G.: ‘ALOHA packet system with and without slots and capture’.
`ARPANET Satellite System .Nate 8 (N 101 1290),June 1972; reprinted in Computer Comm.
`
`7.5 References
`
`
`
`52
`
`
`
`155
`
`om access protocols’, IEEE
`
`_
`
`c
`
`pacity through packet reservation”.
`
`hannel using a control
`pp. C75.3.1—6
`
`
`
`Multimeter: protocols for VSAT networks
`mnbust—
`18 KLIENROCK, L.,
`and LAM, S.S.: ‘Paeket
`
`switching in a multiaccess broadcast
`
`
`channel: Performance evaluation‘, IEEE Trans
`I" 1’3”
`19 CAPETANAKIS,
`
`. Common, April 1975, pp. 410—423
`J.I.: ‘Trcc algorithms
`mls
`Trans. Inf. Thom},
`for packet broadcasting channels’, IEEE
`‘A
`5
`Sept. 1979, pp. 505~51
`
`
`20 MASSEY, J.L.:
`‘Collision resolution 211
`
`
`gorithms and random—access communica—
`
`
`tions’. UCLA Technical Report, ENGSOIG
`, April 1980
`21 GALLAGER, R.C.:
`
`
`‘Conflict resolution
`
`
`Comm.
`in random access broadcast networks’, Pros.
`AFOSR Workshop in
`
`
`Theory and Applications, Sept, 1978, pp. 74—76
`‘22 MASSEYJ
`
`
`
`Ls. ‘ Collision-resolution algorithms and random-ac
`
`
`cess communications’,
`
`in LONGO, 3. 03¢): Multi—use'r commit
`nitation systems. CISM
`265 (Springer
`Courses and Lectures No.
`-Ver1ag,
`
`Vienna, 1931)
`
`23 LIU, T., and TOWSL
`EY, D.: ‘Window and tree protocols for satellite channels’, Prat.
`
`
`IEEE Infowm
`
`, April 1983, pp.
`
`
`24 RAYCHAU
`
`
`DHURI, D
`
`
`.: ‘Announced retransmission rand
`
`
`Tram. Common, Nov. 1
`985, pp. 11834190
`
`
`
`
`25 MERAKOS, L., and K
`
`
`AZAKOS, D.: ‘Multiaccess of a slotted c
`minislot’. Prat. In
`
`
`ti. Goof. cm Communion
`URI
`lions, Boston, June 1983,
`, D.:
`
`
`26 RAYCHAUDH
`
`
`
`ity asynchronous rando
`tlme-oflarrival based coll1s
`
`
`
`
`1011 resolution’. Proc. IEEE Global C
`
`8.6.175
`pp. 4-
`Orleans, Dec. 1985,
`
`
`27 RAYCHAUDHURI, D.:
`‘Sele
`
`
`access protocol’. Int
`
`
`]. Satellite Commun.,
`1989, 1435—447
`28 MUSSER, J., and
`DAIGLE, j.:
`‘Throughput analysis of an asynchronous code
`division multiple acceSS
`(CDMA)
`—7
`system’. Prat. IEEE Intt. Conf on Communications,
`Ph