Fig.1 Some of the effects of GEO and LEO constellations
| SpaceTech Corner | Number 10, Volume II, April 2001 |
A communication satellite functions as an overhead wireless repeater station that provides a microwave communication link between two geographically remote sites. Due to its high altitude, satellite transmissions can cover a wide area over the surface of the earth. Each satellite is equipped with various “transponders” consisting of a transceiver and an antenna tuned to a certain part of the allocated spectrum. The incoming signal is amplified and then rebroadcast on a different frequency. Most satellites simply broadcast whatever they receive, and are referred to as “bent pipes”. The traditional applications were TV broadcasts and voice telephony. Satellite communications for packet data transmissions is being considered. The applications like mobile services, direct broadcast, private networks and high-speed hybrid networks in which services would be carried via integrated satellite-fiber networks are being considered [39]. `
1.2 Asynchronous Transfer Mode (ATM) and Wireless ATM (WATM)
Asynchronous Transfer Mode (ATM) is an International Telecommunication Union- Telecommunication Standardization Sector (ITU-T) standard for cell relay wherein information for multiple service types, such as voice, video, or data, is conveyed in small, fixed-size cells. ATM networks are connection oriented. It is a cell-switching and multiplexing technology that combines the benefits of circuit switching (guaranteed capacity and constant transmission delay) with those of packet switching (flexibility and efficiency for intermittent traffic). It provides scalable bandwidth from a few megabits per second (Mbps) to many gigabits per second (Gbps). Because of its asynchronous nature, ATM is more efficient than synchronous technologies, such as time-division multiplexing (TDM). With TDM, each user is assigned to a time slot, and no other station can send in that time slot. If a station has a lot of data to send, it can send only when its time slot comes up, even if all other time slots are empty. If, however, a station has nothing to transmit when its time slot comes up, the time slot is sent empty and is wasted. Because ATM is asynchronous, time slots are available on demand with information identifying the source of the transmission contained in the header of each ATM cell. ATM transfers information in fixed-size units called cells. Each cell consists of 53 octets, or bytes. The first 5 bytes contain cell-header information, and the remaining 48 contain the “payload” (user information). Small fixed-length cells are well suited to transferring voice and video traffic because such traffic is intolerant of delays that result from having to wait for a large data packet to download, among other things [42].
2.1 Description of the scenarios
2.1.1 Geographically distributed computing
Geographically distributed computing allows more effective resource sharing and improved utilization of computing resources. Major components of this scenario are inter-process communication and remote file I/O systems [37]. The main factor involved this scenario is the distance of separation between communicating nodes and ways to resolve them. The other factor involved is the necessity of broadband communications with QoS guarantees. Satellite communications can solve the distance factor and ATM can solve the requirements of QoS guarantees. The other factors are a big organization’s nature of having geographically dispersed supercomputers and workstations in branch offices and the need to interconnect them. The pre-requisite is the successful interconnection of terrestrial networks in a seamless way.
2.1.1.1 Requirements
The requirements here are QoS guarantee, fast user response, stable
connections, reachability etc.,
2.1.2. Mobility architecture in ATM and WATM networks
In ATM networks, there are different scenarios based on interconnection
of ATM networks (which may be mobile) between themselves and the need to
interconnect ATM end nodes, which may be geographically distributed. This
motivation is on the basis of the following scenarios.
q High-speed network access by ATM end-nodes, which may be portable
(hence mobile).
q A class of applications, with respect to WATM deals with the mobility
of the ATM switch itself. Here pieces of ATM network, each consisting of
ATM switches, could be in motion with respect to the fixed portion of the
network. Application scenarios would involve mobile platforms with number
of users on board. This scenario is pertinent to airplanes, which provides
communication and entertainment services to passengers. Here the ATM end
nodes are not in motion. Another scenario could be that ships (military
and civil) having ATM networks want to communicate among them and with
the land network. The military networks would also entail security features
for intruder-free communication.
2.1.2.1 Requirements
The requirements here are maintaining quality connections, safeguard
QoS guarantees, smooth handoffs, secure communications etc.,
2.1.3 Distance learning and next-generation education
Distance learning and computer aided instructions are very important
and could be
q Broadcast type communications characterized by one-way information
flow
q Interactive communications characterized by full-duplex information
flow and
Self-learning, in which students can retrieve learning materials remotely
[28].
These scenarios require multimedia communications of very high quality and the main hindrance is the distance factor. ATM is the de-facto standard for multimedia communications due to its capacity to guarantee QoS and support for voice, video and data simultaneously.
2.1.3.1 Requirements
The main requirements are QoS guarantees, voice-video synchronization,
large bandwidth, bandwidth on demand, quality multimedia services etc.,
2.1.4 Multimedia and multi-service applications
Multimedia applications like video-conferencing and multi-service applications
(interconnection of circuit-switched and packet-switched networks) scenarios
are classic examples of bandwidth guarantees and bandwidth on demand scenarios
respectively. They also require synchronization over a great distance.
By default, distance is a factor in these application scenarios. Multimedia
communications is also driven by the backbone concept, assumed to be provided
by fiber cables. In many parts, these may be unviable, uneconomic or take
too long to establish. Multi-service communications also entail interconnection
of the mobile devices carried by company representatives.
2.1.4.1 Requirements
QoS guarantees, bandwidth on demand, large bandwidth, synchronization,
and backbone dependability are demanded by multimedia applications. Seamless
and efficient integration schemes are needed by multi-service applications.
Interactive computing and bulk transfers with high bandwidth requirements,
information dissemination including stock market data etc., and video broadcasts
with low delay requirements are some other multimedia applications to be
taken care of [26].
2.1.5 Secure broadband communications
Secure communications are needed by military and sometimes, for big
companies, financial institutions and banks, having distributed branches.
The main factor is that secure communications are needed over a geographically
separated scenario in which distance is the main consideration.
2.1.5.1 Requirements
Security, encryption and interconnection between geographically diverse
locations are the main issues here.
Taking into consideration, all the above factors, SATATMs can be taken as the preferred choice for the above scenarios.
3.1 Constellation of the satellite
The orbital radius of the satellite greatly affects its capabilities and design. The following diagram shows the effects of the constellations for GEO and LEO constellations.
Fig.1 Some of the effects of GEO and LEO constellations
An important measure of efficiency that affects SATATM is end-to-end delay. The end-to-end delay is the sum of transmission delay, uplink delay, downlink delay, ISL propagation delay, OBS/OBP delay and buffering delay. Though LEO networks have relatively smaller propagation delays, the delay variance is higher than GEO. This variation is due to handovers, satellite motion, OBS and adaptive routing. These should be considered while selecting the constellation. Thus, when considering constellation of a satellite, the parameters to be taken into account are launching cost (less for LEO), propagation delay (less for LEO), delay variance (more for LEO, hence bad), coverage (more for GEO, change continuously for LEO), altitude (low for LEO and hence small end-end delays, low power requirements) etc.,
3.2 Handovers and re-routing
GEO systems do not have too many handovers due to its large distance from Earth and due to its high coverage area. Handovers for LEO satellites are estimated to occur on an average 8 to 11 minutes [10]. There is an amount of delay variance in LEO constellation due to these handovers. There are different handover protocols being considered and Footprint Handover Rerouting Protocol (FHRP) is one of them [11]. LEO systems with multi-hop inter-satellite links need handover and rerouting protocols.
Possible after effects of handovers are listed below.
¨ A new satellite may be added to existing connection route
¨ The existing connection route should be updated
¨ A new route/connection must be set up.
Addition of a new node could cause sub-optimal route and hence re-routing
is necessary. This causes additional signaling and processing overhead.
The assumption of FHRP is that all handovers are caused by the mobility
of the LEO satellite instead of the ground terminal.
3.3 Presence of Inter-satellite links
The inter-satellite link is also a part of propagation delay. ISLs may
be in-plane or cross-plane links. In-plane links connect satellites within
the same orbit plane and cross-plane links connect satellites in different
orbit planes. In GEO systems, ISL delays can be assumed to be constant,
while in LEO systems ISL delays depend on the orbital radius, the number
of satellites-per-orbit and inter-orbital distance. The ISL delay in LEO
systems change frequently due to satellite movement and adaptive routing
techniques. Thus LEO systems can exhibit a high variation in ISL delay
[10]. There are some improvements needed to this routing protocol as suggested
in [16] and should be consulted before usage.
Table 1: Relation between traffic models and MAC choices
| Traffic Model | MAC class choices |
| Non-bursty traffic
Bursty traffic Bursty traffic, long messages larges number of users Bursty traffic, long messages, small number of user |
Fixed assignment
Random acces Reservation protocols with contention Reservation protocols with fixed
|
Hence, the usage of ISLs is very much in vogue and recommended and routing strategies to minimize average number of route changes without increase in path delay should be considered before usage. The jitter due to ISLs is also reduced by usage of the routing protocol.
3.4 Presence of OBP/OBS
Some issues to be addressed, before the selection of OBP/OBS are:
o Space environment considerations and associated delays (e.g., GEO
systems)
o Satellite limitations like long transmission delay, link noise, local
weather conditions and interference.
o Cost of operation of satellite and launch costs. The costs associated
with launching satellites with OBS/OBP are high compared to that of bent
pipe satellites.
o Lifetime of the satellite. Generally the satellites last for an average
of ten years.
o Onboard buffer size. This is a very important issue, since the real
estate or memory requirements onboard the satellite are scarce and hence
the buffer size should be carefully chosen. Simulation studies for different
types of ATM traffic are done and should be used [14] before choosing the
value for this parameter.
o Capacity and port rate are other important parameters in addition
to implementation considerations. These are addressed in [33].
o While terrestrial switches should be modular to cater to a broad
range of capacities, OBS could be a lot simpler and tailored to satellite
communications. ?Due to restrictions on payload size and costs, there should
be distribution of ATM-layer functions between onboard switch, NCC and
ground terminals.
o Due to restricted lifetime of satellites, fault tolerance should
be added by introducing fault detection and redundancy, both internal and
external to the switch [33].
o Because of switching delay in the satellites and also to prevent
retransmissions in a long-delay path, the onboard buffers should be larger
than the terrestrial switches to limit onboard congestion.
o Due to hostile radiation environment, particularly in GEO constellations,
the switch ASICs and memory chips for buffers should be suitably safeguarded.
The rad-hard technology is advised [33].
o Switch architectures with a large number of components may be unsuitable
due to satellite limitations in terms of size, mass and power.
o Power consumption and power dissipation are other significant factors
to be considered.
o CLRs should be in the range of 10^-10 to meet the QoS of high-performance
traffic and avoid costly retransmissions [33].
o To get good throughput/delay performance, output or shared queuing
should be used. The output queuing mechanism could be physical buffer based
or virtual buffer based. There are issues in choosing fully output buffered
switch. After sorting through the issues, a fully interconnected fabric
with output port concentrators similar to the knockout switch is being
proposed. The high CLR of these types of switches should also be taken
into consideration [33].
o Functions that could be considered for OBS/OBP are switching, queuing,
flow control and scheduling. Connection admission control and resource
allocation should be handled at NCC preferably. All delay-tolerant functions
should be kept on the ground.
3.5 MAC layer protocols, scheduling and ATM services mapping for QoS
The key difference between a SATATM network and the terrestrial network is the fact that the SATATM network uses multiple access in the uplink. The choice of multiple access schemes has a great impact on the SATATM network. The primary goal in the assignment process is
o Satisfy the user’s QoS in the form of maximum cell transfer delay
(maxCTD), peak cell delay variation (peakCDV) and cell loss rate (CLR).
o Maximize the utilization of the uplink
o Cell delivery in a timely manner and with minimal collisions [26].
Satellite networks present unique challenges in system design related to
QoS provisioning. MAC protocols are behind the delivery of QoS contract.
MAC protocol should achieve QoS provisioning, efficiency and service interoperability
[26].
The traditional CSMA/CD schemes cannot be used with satellite channels,
since it is not possible for earth stations to do carrier sense on the
up-link due to the point-to-point nature of the link. A carrier sense at
the downlink informs the earth stations about potential collisions that
may have occurred 270ms ago. Such delays are not practical [1]. Most SATATM
schemes use dedicated channels in time and/or frequency for each user.
ALOHA, Frequency Division Multiple Access (FDMA), Time Division Multiple
Access (TDMA) and Code Division Multiple Access (CDMA) are such schemes.
The ability to use OBP and multiple spot beams will enable future satellite
to reuse the frequencies many times more than today’s system. Demand Assigned
Multiple Access (DAMA) systems allow the number of channels at any time
is less than the number of potential users. Satellite connections are established
or dropped only when traffic demands them. Protocols like Packet Reservation
Multiple Access (PRMA), an improved form of TDMA with techniques from S-ALOHA,
could also be used.
The following diagram shows two satellite system network scenarios
[26], which can help decide which MAC protocol would be better for different
scenario.
Fig.2 Satellite network scenarios based on traffic aggregation
Demand Assigned Multiple Access MAC protocol can be used, when
o Burstiness of traffic is high
o Low bit-rates are to be supported BW conservation
o Delay requirements not critical.
An in depth study on MAC protocols for Mars Regional Network [47] could be consulted for more information.
Here Tree Contention Resolution Access (Tree CRA) and others are being used.
3.6. Power management
One of the major challenges in the design of a satellite network is the limited transmission power of both the ground terminals and the satellite. Transmissions in the network should be such that the user terminals at different geographical areas are given access in the most power efficient manner [8]. Multi-beam satellites are proposed for this. Multi-beam systems need OBS/OBP. Hence when doing power management, the issues regarding OBP/OBS should also be taken into consideration. To further save on uplink transmission power, MAC protocols like MF-TDMA can be used as the data-link protocol.
3.7 Error Correction Scenarios
In satellite channels under consideration, transmission bit errors occur
in bursts due to link attenuation and use of convolution coding to compensate
for channel noise. Because ATM was designed to be robust with respect to
bit errors randomly distributed, burst errors introduce cell loss (CL).
For a BER of 10^-7, the CL ratio can be as high as 10^-6. Though AAL5 has
a 32-bit CRC, it is not used due to the high cell discard rate at the physical
level [30]. There are several schemes for error correction like interleaving
mechanism, error recovery algorithms, and efficient coding schemes, for
improving error performance.
It has been shown that when interleaving is done, the ATM cell discard
probability (CDP) and probability of undetected errors are less. Interleaving
the ATM cell tends to “distribute” or spread the bit errors at the cost
of increased delay. The interleaving algorithm can be applied differently
according to the AAL types. There is a chance that errors can occur in
the interleaved cells. Another problem is that the interleaving depth for
optimal error performance is still not evident [30].
Error recovery algorithms like automatic repeat request (ARQ) could be used to lower error ratio for loss-sensitive, delay-insensitive scenarios. There are stop-and-wait, Go-Back- N and Selective-repeat algorithms. See [40] for more details in error recovery algorithms. Go-Back-N and Selective-repeat are better than stop-and-wait algorithms. Coding scheme can be used for error correction or prevention. Currently, convolution code with viterbi decoding is used to achieve 10^-3 to 10^5 BER [30]. This is not fit for SATATM networks because of the loss-sensitive ATM traffic. Hence concatenated coding with outer coding as Reed-Solomon (RS) coding with Forward Error Correction (FEC) as the internal convolution code is being currently used and is a good performer in this area [30]. Here also, optimal interleaving depth for SATATM networks should still be determined.
An in-depth study of the impact of transmission error characteristics on SATATMs is studied in [18]. The ATM cell performance measures are Cell acquisition time (CAT), Cell in-synch time (CIT) and cell discard probability (CDP). Satellite links that operate at high rates employ error correction schemes for providing acceptable BER. Burst errors are generated by these error correction schemes. The ATM HEC is capable of correcting only single-bit errors. A method called ATM link enhancement (ALE) was developed, which incorporates a selective interleaving technique allowing it to be transparently introduced into the satellite link. More information is given in the section under Commercial SATATM Products in this paper. Studies confirming its validity are shown in [18]. AAL1 uses a 3-bit CRC, AAL3/4 uses a 10-bit CRC and AAL5 uses 32-bit CRC for error detection and error correction. All the codes used for AALs are sensitive to burst errors, hence the need for better error control algorithms.
In a related experiment [47], an error correction scheme using side information is proposed to improve the throughput of ATM transmission over Rayleigh fading channel like a satellite link using binary phase shift keying (BPSK) modulation. The method combines the ARQ protocol and the error correction scheme with throughput.
In another experiment [24], a shorter error correction model called Bose-Chaudhuri-Hocquenghem (BCH) code could be used. A more ATM oriented solution is also discussed, which is called the Partial Packet Discard (PPD), which on detection of erroneous cells at the satellite switch, these and consecutive ones are dropped and hence reduce the traffic. This suffers from the retransmission problems (increase in congestion) due to obvious reasons. The study goes on to explain implementations for the different AAL layers for ATM. A comparison of PPD approach with a LLC layer mechanism is also carried out.
3.8 Traffic Control and Congestion Control
Traffic Control is a measure that takes actions to avoid congestion conditions. Congestion control acts after congestion is set. Traffic Control is congestion avoidance. This is very important, since the satellite links are bandwidth limited [30]. The algorithms should act faster and more efficiently due to the long delay. The basic QoS parameters are Cell Loss Ratio (CLR), maximum and mean cell transfer delay (CTD) and cell delay variation (CDV) and the extended QoS parameters are cell error ratio (CER), severely errored cell block ratio (SECBR) and cell mis-insertion ratio (CMR) are also recommended. The impact of satellite delay on some basic services is tabulated here[30].
Table 2: Performance comparison
| Acces
Protocol |
Efficiency | Delay | Satellite | Robustness | Complexity |
| S-ALOHA | 0.37 | Low | Low | High | Low |
| Tree-CRA | 0.43-0.49 | Medium | Medium | Poor | Medium |
| DAMA
(reservation) |
0.6-0.8 | High | High | High | Medium |
| Hybrid
(reservation random) |
0.6-0.8 | variable | Medium | High | Medium |
Congestion control mechanisms are of many types [30]. A frequently used scheme is selective cell discard. It has advantages and disadvantages as briefed above. Another method is Explicit Forward Congestion Indication (EFCI) incorporated with a feedback mechanism. EFCI is used to convey congestion notification to the source. The destination protocol is required to notify the source of congestion. This whole process is Forward Explicit Congestion Notification (FECN). In SATATM, this is not very well matched due to the minimum delay of one-way propagation for the notification. Backward Explicit Congestion Notification (BECN) is a mechanism, which could be used to send a notification in the reverse direction of the congested path. Buffering and VC prioritization can also be used in congestion control. The satellite on-board buffer could also be considered. This could introduce jitter, if not properly done. A related mechanism is VC prioritization. Other congestion control mechanisms are discussed in [49], although these should be changed for SATATM considerations. Thus the criteria of choosing the algorithms should be that, these should not affect the delay-sensitive traffic for SATATMs.
3.9 Upper layer concerns
There is a service specific convergence sub-layer (SSCS) in AAL. This SSCS is divided into service co-ordination function (SSCF) and SSCOP. The service specific connection oriented protocol (SSCOP) can run on all protocol stacks. Its main function is to provide assured delivery of PDUs and use error-recovery procedures if necessary. The following features [18] are very favorable to SATATM networks. They are the selective retransmissions, nearly infinite window size definition capability, superior flow control, optimized support for high-speed and long-delay networks and the protocol is designed to be insensitive to network delay. SSCOP has been proposed by some people as a possible replacement for TCP as a wide-area transport protocol, however some doubts have been expressed as to its efficiency in the face of errors, congestion, variable delays. A thorough investigation of SSCOP, including simulation to determine its performance in terms of throughput etc., in a typical error/congestion/delay environment should be carried out. TCP is the de-facto standard for the Internet transport protocol. Considerations for using TCP over ATM over satellite communications have been studied in sufficient depth [5,6,7,20].
o Maximum window size remains a hindrance to the SATATM networks. An
increase in the window to 2 ^30 is being proposed [6]. Also, larger initial
window size has been recommended
o TCP for transactions could be used due to the lesser number of handshakes
o Slow start wastes network capacity and are also inefficient for transfers
that are shorter in size.
o To counter delayed ACK caused delay in the sender side to increase
the window, “byte counting” approach is being studied. Otherwise, delayed
ACKs must be used only after the slow start phase.
o The Fast Recovery method should take into account, information provided
by SACKs sent by the receiver.
o The Forward Acknowledgement algorithm was developed to improve TCP
congestion control during loss recovery.
o Explicit congestion notifications should be used
o Differentiating between congestion and corruption is a difficult
problem for TCP. Doing it would be of great use to TCP over SATATM networks.
This is handled in [20].
o During congestion avoidance, in the absence of loss, the TCP sender
adds approximately one segment to its congestion window during each RTT.
This leads to unfairness and hence fair queuing and TCP-friendly buffer
management in network routers is being considered.
o The use of multiple data connections for transferring a file in a
SATATM network impacts the network and should be used after careful review.
o Rate-based pacing (RBP) is being considered to counter the slow window
opening during slow-start and could be used.
o TCP header compression is a viable alternative for bandwidth-sensitive
SAT networks.
o Sharing TCP state among similar connections could be used to overcome
limitations in the configuration of the initial state.
o In highly asymmetric networks like satellite links, a low-speed return
link can cause performance drop due to congestion in the acks returning
to the sender. Hence Ack Congestion Control (ACC) must be done.
o Ack filtering can be done in the previous case to limit the number
of acks in the return direction. This could be done taking into advantage,
the cumulative acknowledgement scheme of TCP. These are some of the TCP
improvements to be made for supporting satellite networks in general and
will apply to SATATMs as well.
3.10 Attenuation considerations
Path loss can occur in satellite transmissions due to the following conditions. Weather conditions like rain, integrated water vapor concentrations and cloud liquid water contents can affect the transmission. Attenuation due to rain is a major problem in the Ka and Ku bands. The effect of airline traffic on satellite transmissions is also studied [45]. Global predictions of slant path attenuation are also being studied and should be taken into consideration. Information related to attenuations could be further studied at [54].
3.11 ATM layer changes for satellite considerations
There are changes proposed to the ATM layer and specifically in the
Service Specific Sub-layer of AAL to incorporate satellite communications
as the physical layer transport of ATM. In one type of change, the CRC
and the sequence numbers are moved to the higher convergence sub-layer.
This will entail larger blocks to put error detection and correction on.
It is suggested here [9] that high-speed, long-delay satellite links need
a unique AAL. That is answered in [24], where a separate layer called S-ATM
layer is provided for satellite communication scenarios. In another study
[43], the ground segment proposal is based on a new AAL called AAL2, which
is considered to play a major role in offering an efficient way to provide
multimedia services over ATM networks. It allows easy encapsulation of
the complete set of media component sessions, which forms a multimedia
transaction into a single ATM VC connection.
Other aspects such as link budget scenario, elevation angles, and encryption of traffic, should also be considered. Because of limited spaces, interested readers can see the materials in references.
Srihari Raghavan (sraghava@vt.edu)
Virginia Polytechnic Institute and State University