Академический Документы
Профессиональный Документы
Культура Документы
relatively small number of large earth stations which have a relatively large average bit rate. The cost and
the size of the earth station has a small impact on the suitability of the satellite solution.
The RACE CATALYST project was a demonstrator for this scenario and showed the compability of satellite
technology with ATM and the terrestrial B-ISDN. The equipment developed during the CATALYST project
has been able to interconnect ATM testbeds as well as the existing networks such as DQDB, FDDI and
Ethernet networks, all using ATM. A detailed explanation of the system design and performance is provided
in [POLE94], [LOUV94], [HADJ94], [SUN95].
In the second scenario the satellite system is located at the border of the B-ISDN and provides access links
to a large number of users. This scenario is characterized by a large number of earth stations whose average
and peak bit rates are limited. The traffic at the earth station is expected to show large fluctuations.
Therefore the multiple access scheme will considerably effect the performance of the system. Furthermore
the cost and size of the earth station have a large impact on the suitability of the satellite solution.
Very Small Aperture Terminal (VSAT) satellite systems could be used for the second scenario. The objective
of this paper is to develop a system that uses on-board processing (OBP) satellites and VSATs in order to
make B-ISDN access affordable for a large number of users by lowering earth-station cost and providing
bandwidth on demand. The problems that need to be resolved are investigated.
2. ATM VSAT SATELLITE SYSTEM ARCHITECTURE
2.1 ATM Networking via Satellite
Computer networking has seen a tremendous growth during the last two decades, and there are today
numerous global networks, using a mixture of terrestrial and satellite links. However, there is little
experience concerning the interconnection of broadband networks using high-speed satellite links. The
problem is made difficult from the fact that such a network must cope efficiently with both the high
transmission rates of ATM/B-ISDN and the substantial satellite channel round-trip propagation delay.
One of the major issues seen as affecting the rapid implementation of B-ISDN over the coming decade is
the requirement to provide such a service over a wide geographical area, which is not possible without a
substantial initial investment. The actual demand and traffic characteristics of potential network users are
yet unknown, and future customers are likely to be skeptical about joining a network at a very high initial
premium. Communication satellites, as a possible way of offering access to the broadband network appear
to be a very attractive option because:
ATM services can be provided rapidly over a large area, without the need of major investment.
Satellite communication systems can be complementary to terrestrial networks, especially for
widely dispersed users.
The broadcast nature of satellites can be used where the same message has to be send to a large
number of stations (point-to-multipoint transmissions).
A wide range of customer bit rates and circuit provision modes can be supported.
New users can be accomodated swiftly with simply installing new earth stations at customer
premises (cost of earth station will have an important impact). Thus possible network enlargement
is not a significant planning problem.
The network architecture consists of a large number of VSAT earth stations which are interconnected in a
mesh configuration by an cell switching on-board processing (OBP) satellite with spot beams.
37/2
capabilities.
The major issues in developing an OBP cell switch include: (i) the traffic model, (ii) traffic and congestion
control, (iii) testing and evaluation. Section 3 explains the used traffic model and traffic/congestion control
mechanism in detail whereas Section 4 provides simulation and theoretical results of the system
performance.
3. TRAFFIC AND CONGESTION CONTROL
Providing the desired QoS for various traffic categories in ATM networks is not an easy task. The design
of a suitable ATM traffic and congestion control is the most important challenge for the success of an ATM
based B-ISDN. Therefore it has been the subject of vigorous research over recent years.
Various control mechanisms have been proposed for ATM networks. These can be classified into two
categories: reactive control and preventive controls. Preventive control techniques attempt to prevent
congestion by taking appropriate actions before they actually occur. A preventive flow control mechanism
consists of the connection admission control (call level) and usage parameter control (cell level). Reactive
control (burst level) is a technique used to recover from a congested state.
This Section will first introduce the ATM service categories. Then more detailed information about the used
connection admission control, usage parameter control and reactive control mechanism is provided.
3.1 Service Categories
According to [ATMF95] services provided at the ATM layer consist of the following five service categories:
CBR
rt-VBR
nrt-VBR
UBR
ABR
Service categories are distinguished as being either real-time (rt) or non-real-time (nrt). CBR and rt-VBR
are real-time categories whereas nrt-VBR, UBR and ABR are non-real time categories.
3.1.1 Constant Bit Rate (CBR) Services
CBR services generate traffic at a constant rate and can be simply described by their peak cell rate (PCR).
The burstiness of a CBR source is equal to one, and the source is active during the duration of the
connection (or the silent periods are also transmitted at the peak rate). This service category is intended for
real-time applications, i.e. those requiring tightly constrained delay and delay variation, as would be
appropriate for voice and video applications. Cells which are delayed beyond the value specified by Cell
Transfer Delay (CTD) are assumed to be significantly reduced value to the application and might be
discarded.
3.1.2 Variable Bit Rate (VBR) Services
The traffic generated by a typical VBR source, in general, either alternates between the active and silent
periods and/or has a varying bit rate generated continuously. The peak-to-average bit rate (burstiness) of a
VBR source is often much greater than one. VBR services can be described by different sets of traffic
descriptors. The used TDs are described in Section 3.2.1.
The rt-VBR service category is intended for real-time applications such as voice and video guaranteeing a
certain CTD. The non-real-time category is intended for non-real time applications which have bursty traffic
characteristics and not so stringent delay characteristics. A bound on the mean transfer delay is however
37/4
provided. For both rt- and nrt-VBR services the required low cell loss ratio (CLR) is guaranteed by the
network.
3.1.3 Unspecified Bit Rate (UBR) Services
The UBR service category is intended for non-real-time applications. This category does not specify traffic
related service guarantees. Specifically, UBR does not include the notion of a per-connection negotiated
bandwidth. No numerical commitments are made with respect to the cell loss rate experienced by a UBR
connection, or as to the cell transfer delay experienced by cells on the connection.
3.1.4 Available Bit Rate (ABR) Services
ABR is a service category for which the limiting transfer characteristics (such as PCR) provided by the
network may be changed subsequent to connection establishment. A flow control mechanism is specified
which supports several types of feedback to control the source rate in response to changing ATM layer
transfer characteristics. If the end-system adapts its traffic in accordance with the feedback it is guaranteed
a certain CLR and it will obtain a fair-share of the available bandwidth according to a network specific
allocation policy. ABR service is not intended to support real-time applications.
On establishment of an ABR connection, the end system specifies both a maximum required bandwidth
(PCR) and a minimum usable bandwidth (MCR) to the network. The MCR may be specified as zero. The
bandwidth available from the network may vary, but shall not become less than MCR.
Only the rt-VBR and ABR traffic categories are considered since they represent integrated transport of realtime and non-real time traffic flows. CBR traffic and UBR traffic can easily be integrated in the traffic
management scheme. The bandwidth assignment to CBR services is the peak rate and traffic control of the
peak rate is not difficult. Thus CBR traffic will only have the effect of background traffic. Handling UBR
traffic introduces the issue of protecting the QoS objectives of the ABR connections.
3.2 Connection Admission Control (CAC)
The CAC decides whether or not a connnection can be accepted. Network resources are allocated according
to the traffic contract which is negotiated between the user and the network. Parameters which form this
contract are the Traffic Descriptors, QoS requirements, Conformance Definition and the Service Category.
A connection request is accepted only when sufficient resources are available to establish the call through
the whole network at its required QoS and maintain the agreed QoS of existing calls. This also applies to
re-negotiation of connection parameters within a given call.
There are two alternative approaches for bandwidth multiplexing: deterministic or statistical multiplexing.
In deterministic multiplexing each connection is allocated its peak bandwidth. Although this can eliminate
cell level congestion, it goes against the philosophy of ATM, which offers the potential for improved
bandwidth utilization through statistical multiplexing of variable bit rate and bursty traffic. There is a need,
specially for satellite links which are bandwidth-limited compared to optical fibre links, to fully use the
statistical multiplexing capabilities of ATM.
3.2.1 Traffic Descriptors and Parameters
The Connection Traffic Descriptors (CTD) play an important role in the preventive control scheme. The
Connection Admission Control (CAC) has to consider the CTD in order to allocate the necessary network
resources for the connection. The Usage Parameter Control (UPC) then monitors the conformance of cells
to the negotiated CTD. This ensures that the unintentional or malicious behaviour of some users will not
result in a performance degradation for other users.
An important issue is the set of traffic parameters to be included in the CTD. Only the peak cell rate and
Cell Delay Variation (CDV) tolerance has been standardized by the ITU-T. Some other widely used traffic
37/5
parameters have also been proposed: mean rate and mean burst length [VAKI91], sustainable cell rate and
burst tolerance [ATMF94]. In this paper we will use the peak rate, CDV, mean cell rate and mean burst
duration as traffic parameters.
3.2.2 Source Characterization and Source Model
The traffic model, which is widely used for the characterization of ATM sources, is the on-off source model.
This model has been successfully used to realistically model packetized speech, still picture and interactive
data services [ONVU94]. Using the on-off model the ATM cell stream from a single source is modelled as
a sequence of alternating burst periods and silence periods as shown in Figure-4. a and b are the transition
rates between the on and off states. The duration of each burst is exponentially distributed with mean
1/a ms. During such a period ATM cells are transmitted with constant interarrival time T ms, where T =
1/PCR. After generation of the ATM cells an exponentially distributed silence period with mean value 1/b
ms follows. If N identical on-off sources are multiplexed, [WEIN78] found that the number of active sources
i can be modelled very well by a continous-time birth-death (b-d) proccess, shown in Figure-5. The
interrarival time of cells in state i is 1/(i PCR). This b-d model greatly simplifies the simulation by reducing
the simulation time since for N sources 2N states are required for the on-off model and N states for the b-d
model. The parameters used by both models governing the transition rates are: mean burst duration = a-1,
mean silence duration = b-1.
3.2.3 CAC Algorithms
A variety of CAC algorithms have been proposed. The aim is to use an algorithm that is simple (in terms
of processing and storage requirements) and efficient (to allow statistical multiplexing gain).
We use a modified binomial model. The binomial formula was used in [SYKA92] with success, as the buffer
size to burst length ratio, for the source types considered there, was close to zero. The maximum number
of identical sources in the on-state without cell loss is equal to the capacity C=link rate/PCR. As the number
of active sources becomes larger than C, the output link is not able to carry the required bandwidth (bw)
and cell loss occurs. Our modification in the binomial model is that we assume that cell loss will occur in
case of an overload situation where i > C+2.
This assumption is justified when many connections are multiplexed and the buffer size is large enough,
compared to the burst length, to absorb one source in temporary overload. The equilibrium probability of
i sources being active Pbi is given by the binomial distribution
(1)
Dividing the rate of lost ATM cells with the maximum number of generated cells N1PCR, the cell loss rate
(CLR) is obtained:
(2)
where
1=a
-1
/(a-1+b-1).
established connections.
The Leaky Bucket (LB) is generally agreed to achieve the best performance compromise of the mechanisms
studied for UPC [ONVU94] [CHEN95]. It was first introduced in [TURN86]. Since then a number of
variants have been proposed. The basic idea behind this approach is that each incoming cell needs a token
to enter the network. Tokens are generated at constant rate r. The size of the bucket imposes an upper bound
on the burst length and determines the number of cells that can be transmitted back to back, controlling the
burst length. Provided that the burst is short, the bucket will not empty and no action will be taken against
the cell stream. However, if a long burst of higher-rate cells arrives, the bucket will empty and the UPC
function will take actions against cells in that burst. The tolerance allowed for the connection depends on
the size of the token buffer (M) and the token generation rate (r) which are also the parameters of the LB.
Conceptually, the tokens can be viewed as arrivals to a finite-capacity, single-server queue with deterministic
service time. It is also obvious that the LB enforces the rate r and allows temporary bursts above the rate
r depending on the bucket size (M). The implementation requires a simple up/down counter to reflect the
contents of the token bucket.
A number of variations of the basic LB are possible. Instead of discarding or marking cells when the token
bucket is full, arriving cells can be allowed to queue in a data buffer (BD). The input or data buffer smooths
the burst by spacing the cells at the cost of introducing some delay.
The two types of enforcement action that can be taken within the LB scheme (cell discarding or marking)
, and whether or not there is a user buffer, gives rise to different versions of the LB.
3.3.1 Control of the peak cell rate
The peak rate of the source is effectively controlled by setting the token generation rate near the peak cell
rate and allowing some margin for the CDV. Previous work [WILT94] has analyzed the CDV caused by
various topologies in the access network and has given some typical values that could be used as default for
the buffer size required to achieve a CLP within a CDV tolerance.
3.3.2 Control of the mean cell rate
The mean cell rate can be very easily controlled if the burst tolerance (maximum burst size) is known. Then
the dimensioning of the LB is very straightforward. However if only the mean burst duration is known than
the mean cell rate is not so easy to control. This is due to the fact that a long observation time is required
before detecting any non- conforming cells. A large token buffer has therefore to be chosen in order to
minimize the probability of dropping/marking cells which conform to the negotiated traffic parameters. This
however causes a long reaction time to increases in the mean cell rate whereas fast detection of violations
is required. This problem can be solved by introducing a data buffer (B D). In order to distribute the buffer
to obtain the best performance, in terms of reaction time and queuing delay, the buffered LB was analyzed
in [ORS95].
3.4 Reactive Congestion Control
Although preventive control tries to prevent congestion before it actually occurs the satellite system may
experience congestion due to multiplexing buffer or switch output buffer overflow. In this case, where the
network relies only on the UPC and no feedback information is exchanged between the network and the
source, no action can be taken once congestion has occurred.
Many applications, mainly handling data transfer, have the ability to reduce their sending rate if the network
requires them to do so. Likewise, they may wish to increase their sending rate if there is extra bandwidth
available within the network. This kind of applications are supported by the ABR service class. The
bandwidth allocated for such applications is dependent on the congestion state of the network. Rate-based
control was recommended for ABR services, where information about the state of the network is conveyed
to the source through special control cells called Resource Management (RM) cells [ATMF94]. Rate
37/7
The CLR experienced by the ABR traffic was observed as very low for a satellite network utilization of 95%
and a multiplexing buffer size of 100 cells. The reson for chosing such a high utilization is the fact that the
binomial model slightly overestimates the cell loss, assigning too much bandwidth to a source. The MCR
for each ABR source was assumed zero so that the bandwidth assigned to each source was the same. Further
simulations are required to obtain the CLR of ABR traffic on-board the satellite under different load
conditions.
5. CONCLUSIONS
As user demand is becoming more complex, VSAT satellite networks, which have so far been successfully used in
providing specific communication services are expected to provide a much wider range of services (such as multimedia).
An on-board cell swotching satellite with spot beams is considered for the scenario where a high number of VSATs
are interconnected in a mesh configuration. In our proposal maximum utilization of the burst slots using complex
DAMA schemes, which introduce high delay, is not required. Maximum bandwidth utilization of the satellite is achieved
by statistical multiplexing the traffic on-board the satellite. Greater connectivity, improved spectral efficiency, improved
use of the switch capacity and greater e.i.r.p. per VSAT terminal are the main advantages of using a state-of-the-art
satellite system. These advantages enable new services while potentially reducing the cost of earth stations.
However development of an OBP cell switched satellite communication network introduces some issues like traffic and
congestion control which have been addressed in this paper. It has been shown that rt and non-rt traffic can be
integrated while still achieving increses in throughput up to 228% for a very low CLR.
We conclude that the use of VSATs for ATM services is possible, but careful system design and dimensioning is very
important to provide the required quality of service.
37/9
UNI
UPC
VBR
Source
UPC
RF
.
.
Cell
CBR
Source
Mod/
Demod
Interleaver
Scheduling
and Processing
UPC
ABR
Source
VBR
Source
Control
Module
UPC
.
.
ABR
Source
UPC
Delay
Explicit Rate
Feedback
Signal
UBR
Source
UPC
.
.
UBR
Source
on-off
birth-death10
birth-death20
Theoretical
0.10
0.05
UPC
0.00
0.0
0.2
0.4
0.6
0.8
Number of active sources / 50
1.0
control of
the mean rate
Data Buffer
to multiplexer
0.12
No of cells in state i / Total no of gen. cells
very small
Token buffer
Token Buffer
token generation rate
near peak rate
token generation rate
near mean rate
0.08
0.06
0.04
0.02
0.00
0.0
2 ATM Cells
106 bytes
4 bytes
50 voice sources
100 voice sources
400 voice sources
800 voice sources
0.10
0.1
0.2
0.9
Synchronization
and Phase
Header
Resolution
Figure-3 Frame Unit Format
10
10
10
a
10
1-b
CLR
1-a
Burst
10
10
Silence
10
10
2b
(N-1)b
...
2a
10
(N-1)a
N=400
N=800
10
0.35
0.40
0.45
Capacity / N
N-1
Na
1.0
REFERENCES
[ATMF94] ATM-Forum, ATM User-Network Interface Specification Version 3.1, September 1994
[ATMF95] ATM-Forum, Traffic Management Specification Version 4.0, 95-0013R8, October 1995
[BARN95] BARNHART A.W., "Baseline Model for rate-control simulations", ATM Forum 940399,Sept.1994
[CHEN95] CHEN T.M. and LIU S.S., ATM Switching Systems, Artech House, 1995
[FROM92] FROMM H.H. and BELLA L., "System architecture and design criteria for the ESA OBP
development",pp.57-62, 9th Int. Conf. on Digital Sat. Com., Denmark, 1992
[HADJ94] HADJITHEODOSIOU M.H., KOMISARCZUK P., COAKLEY F., EVANS B., and SMYTHE
C., "Broadband Island Interconnection via Satellite-Performance Analysis for the RACE II-CATALYST
Project", International Journal of Satellite Communications, V.12, pp.223-238, 1994
[HADJ96] HADJITHEODOSIOU M.H. and COAKLEY F.P., "Networking with VSATs: From low bit rate
systems to ATM service provision", AIAA 16th International Comms Satellite Systems Conference, pp.
854-864, 1996
[IVAN94] IVANCIC W.D., CHU P., and SHYY D., "Technical Issues regarding Satellite Packet Switching",
International Journal of Satellite Communications, V.12, pp.257-267, 1994
[JAIN95] JAIN R., "Congestion Control and Traffic Management in ATM Networks: Recent Advances and
A Survey", Computer Networks and ISDN Systems, February 1995
[LOUV94] LOUVET B. and CHELLINGSWORTH S., "Satellite integration into broadband networks",
Electrical Communications, 1994
[ONVU94] ONVURAL Raif O. Asynchronous Transfer Mode Networks: Performance Issues, Artech House,
1994
[ORS94] ORS T., An Investigation of the Leaky-Bucket for Preventive Control in ATM Networks, Masters
thesis, University of Surrey, September 1994
[ORS95] ORS T. and JONES S.P.W., "Performance Optimization of ATM Input Control using Multiple
Leaky-Buckets", Third IFIP Workshop on Performance Modelling and Evaluation of ATM Networks, July
1995
[POLE94] POLESE P, MORT R., and COMBAREL L., "Satellites in UMTS and B-ISDN: status of
activities and perspectives", Electronics and Communication Engineering Journal, pp. 297-303, December
1994
[RACE93] RACE Common Functional Specifications D751, "Satellites in the B-ISDN, General Aspects",
Issue D, December 1993
[SUN95] SUN Z., COAKLEY F.P., and EVANS B.G., "Satellite ATM for Broadband ISDN
Telecommunication Sytems", V.4, pp.119-131, 1995
[SYKA92] SYKAS E.D., PASCHALIDIS I.Ch., and VLAKOS K.M., "Congestion Avoidance in ATM
Networks", IEEE INFOCOM, 1992
[TUCK88] TUCKER R.C.F., "Accurate Method for analysis of a packet-speech multiplexer with limited
delay", IEEE Transactions on Communications, V.36 No.4, pp.479-483, April 1988
[TURN86] TURNER J.S., "New Directions in Communicationss (or Which Way to the Information Age ?)",
IEEE Communications Magazine, October 1986
[VAKI91] VAKIL F. and SAITO H., "On Congestion Control in ATM Networks", IEEE LTS, pp55-65,
August 1991
[WEIN78] WEINSTEIN C.J., "Fractional speech loss and talker activity model for TASI and for packetswitched speech", IEEE Trans. on Communications, V.COM-26, pp.1253-1257, 1979
[WILT94] WILTS R., WITTERS J., and PETE G.H., "Throughput Analysis of a Usage Parameter Control
Function Monitoring Misbehaving Constant Bit Rate Sources", Second IFIP Workshop on Performance
Modelling and Evaluation of ATM Networks, 1994
[WRIG92] WRIGHT D. and BALOMBIN J., "ACTS System Capabilities and Performance", pp.1135-1146,
AIAA92
37/11