Before
the discussion of wireless ATM can begin, the concept of ATM in general needs
to be discussed. ATM, or by its more
formal name Asynchronous Transfer Mode, is a basic packet-based networking
system designed for the simultaneous transmissions of voice, video, and
data. In the mid 1980s, the major
telecommunication companies decided that they needed a new network to handle
the surge of video and data, along with voice, traffic being sent over their
existing networks. From this, the
concept of ATM was born. From an
increasing need to handle data traffic, which is inherently packet-based, as
well as voice traffic, ATM was designed to work as a packet-switched
network. In a packet-switched network,
all traffic is broken into small pieces, which are easier to transmit than one
large chunk of data. The problem with
using this type of network design for ATM is that the old telephone network is
circuit-switched, or in other words creates a physical direct connection
between the source and destination during the transmission. ATM, therefore, is designed so that it can
handle circuit-switched traffic on its packet-switched backbone. To accomplish this, ATM creates virtual
circuit connections over the packet-based network between the source and the
destination. These virtual circuit
connections provision a set number of network resources dedicated to the
connection between a specific source and destination, making it appear to the
old telephone network that a circuit connection is established. This allows an ATM network to guarantee the
same or greater quality of service for voice traffic as the old telephone
network does, while at the same time providing a much greater level of service
for data and video traffic than was previously available.2
The idea of breaking voice into packets created quite a problem for the standards process. The US telecom carriers wanted to set the packet size to 64 bytes, which is the size of one voice data packet. The European telecom carriers wanted to set the packet size to 32 bytes so that their transmission lines would not require echo cancellation equipment. Instead of making a technical compromise, a more radical solution was implemented. The two proposals were averaged together to get the packet size. The standard was set to 48 bytes with a 5-byte header, creating a 53-byte ATM cell, or packet. Since 48 is not a power of 2, it is completely out of character for normal data standards. The final data packet size of 53 bytes is technically imperfect, but is still useful for all of the types of traffic that would need to be transmitted over an ATM network. 3
In
order for ATM networks to be useful to the telecoms, they needed to be able to
transmit large amounts of data. ATM was
defined to allow data rates ranging from 25 to 622 Mbps (megabits/sec). To achieve these data rates, ATM is designed
to assume that no packets or a very minimal number of packets will be lost
during transmission, so that retransmission can be avoided. This is accomplished by requiring that ATM
networks use fiber optic cable as the transmission medium. Fiber optic cable is the only medium that
has a low enough transmission error rate for the small 53-byte ATM cells to be transmitted
at a frequency fast enough to achieve 622Mbps. 3
With the technical capabilities outlined above, it is
possible to understand why there is such an effort to create wireless ATM
networks. These networks would allow
for the rapid transmission of all types of traffic without wires while using
existing ATM equipment. However, even
though ATM was developed by the telecom companies with all of the technical
characteristics mentioned above, ATM networks have begun to fall out of favor
in the last few years. Instead of ATM,
networks are being constructed using mostly standard IP (Internet Protocol) based
equipment because it has become both cheaper and faster than ATM. Even though ATM may be technically superior
for voice, IP data traffic has become much more important to the telecom
companies in terms of revenues and amounts of traffic, so ATM appears to be a
dying breed.
Wireless
ATM is mainly considered as an “access to an ATM network” issue. Different types of wireless networking need
to be addressed depending on what kind of ATM network is accessed. The proposed technology provides an
extension of the network to mobile users, simplifies wiring and
reconfiguration. When making the
transition to wireless access, there are a number of things that need to be
considered. Consideration needs to
given to the physical, data, and the multi-access layer issues. 1
The
physical layer deals with the transmission of data over the physical medium by
means of a radio or an optical transmitter/receiver pair. Circuit switched or packet switched
operations, operating frequency, licensed vs. unlicensed bands, channel coding,
and the need for multiple antennas are only some of the issues in consideration
when implementing the physical layer for wireless ATM.
·
Radio is the preferred solution since it is not restricted
to line of sight, does not require pointing, and multi-access is simpler to
attain. Infrared is an option but it
mostly operates as a collection of point-to-point communication links. Overcoming multi-path reception from
stationary and moving objects is the main challenge of this endeavor because
the above-mentioned produces space- and time- varying dispersive channels.
·
A pure circuit-switched system is not viable for wireless
ATM due to the presence of variable bit rate services and a hybrid of circuit
and packet – switched systems makes the best alternative. The trade-off,
however, is that it makes the implementation more complicated.
·
When dealing with frequency, the ideal frequency and whether
it is licensed or unlicensed needs to be decided upon. Licensed bands require FCC approval, which
is a long and difficult process.
·
To spread the transmission over a larger number of wireless
channels, wireless ATM implements OFDM (Orthogonal Frequency Division
Multiplexing). OFDM chooses wireless channels that are orthogonal to each
other in the electromagnetic spectrum, which means that they
are completely independent from each other. This minimizes
interference and cross talk, while at the same time maximizing bandwidth.
·
As far as channel coding is concerned, there is no consensus
on the feasibility of channel coding for wireless channels. Due to the busty nature, it either has
errors or it is flawless. System
performance improves by incorporating error correction in unconventional
ways. By replacing the CRC (Cyclic
Redundancy Check) in the data link layer with an error detecting and correct
code, the advantages of error correction can be obtained at most with a slight
increase in the CRC field size.
·
Multiple antennas implemented with the proper selection
algorithm provide for the best transmission performance, but the number needed,
that will provide optimal performance, must be decided. 1
Encapsulation,
header compression, and ARQ vs. FEC are some of the many issues involved in the
data link layer implementation.
·
Encapsulation is a technique used for transporting data
units of a protocol within those of another.
This technique provides the advantage of transparency, but the
disadvantages are added overhead and delays due to encapsulation and
decapsulation.
·
In header compression, the information content of a header
is represented with a fewer number of bits than the 40 bits used in
conventional ATM in order to compensate for the intolerable 10% overhead that
normally occurs.
·
Various ARQ (Automatic Repeat Request) and FEC (Forward
Error Correction) techniques have been compared for various channels. In studying these, the consensus reached is
that the best technique is a combination of ARQ and FEC. Although there are several hybrids currently
present, new research results under the newly assumed conditions.1
In transitioning from ATM to wireless, two major
issues arise. One is the shared use of
unreliable transmission links, and the other is the mobility of the
terminals. The multi-access (MAC)
protocols attempt to efficiently and equitably allocate use of a communications
channel to independent, competing users.
Many schemes have been implemented over the last few decades and each
has its advantages and limitations.
When transitioning to wireless, it necessitates the need to consider one
of these numerous protocols. Mobility
management is another issue that needs consideration in this transition because
the following three basic issues arise:
location management, connection management, and handoff management. 1
There are many advantages of having a wireless ATM
network. The first and foremost major
advantage is that all of the benefits of ATM are made mobile. There is the capability of having a
high-speed network, such as ATM, in the palm of your hand. Therefore, the use of mobile equipment to
communicate and exchange multimedia traffic over small, handheld portable
equipment is one of the options that wireless ATM proposes to offer. Another advantage is flexible bandwidth
allocation that would provide end-to-end communications in a Wide Area
Network. Companies would no longer have
to buy extra equipment like routers to interconnect their LANs (Local Area
Networks). 4
Wireless ATM efficiently multiplexes traffic from bursty
data and multimedia sources. It chooses
signals to avoid congestion and maintain a constant flow of information sent to
the stations. The basic structure of
wireless ATM is that there are a large number of small transmission cells,
called pico cells, that are each served by a base station. All base stations in the network are
connected through the wired ATM network.
The use of ATM switching for inter-cell traffic avoids developing a new
backbone network with sufficient throughput to support communication among the
large number of cells. The basic role
of a base station is to act as an interconnection between the LAN and the
wireless subnets and to transfer packets to the wired ATM network from the
mobile units. Each wireless base
station has two virtual circuits open to each base station and to each
router. As the packet arrives to the
base station of the wireless unit, it chooses the circuit that leads to the
correct destination. Using two
connections guarantees that routing information will not be confused with data
because data packets never travel on the virtual circuits used for routing, and
routing packets never travel on the circuits used for data. The circuits can also be assigned priority
to guarantee that stations receive and process routing updates quickly. Since ATM switching equipment for inter-cell
switching is available for wireless ATM, the availability of using existent ATM
switching is another advantage. 4
The wireless scheme used in wireless ATM allows for the
reuse of spectrum. In other words, to
avoid needing to stop and restart transmitting when moving between pico cells,
the base stations can operate on the same frequency. Small cell sizes give flexibility, thus avoiding the problem of
running out of bandwidth. Along with
this reuse of spectrum, the network is capable of soft handoff without any data
loss. Handoff is when a mobile unit
leaves the area of one cell and enters the area of another. Therefore, soft handoff without any data
loss is important and implemented transparently. With wireless ATM arises the capability of being able to roam
freely. Mobile units are allowed to
roam freely from cell to cell without any user intervention. 4
The major disadvantage of wireless ATM is that it brings
delay to multi-path interference.
Reducing the size of pico cells (transmission cells) brings delay to the
multi-path network as well as a lack-of-sight path resulting in high
attenuation. There are a small number
of mobiles within the range of any base station. As cell size is reduced, hand-over rate increases. By using the
same frequency, there is no hand-over required at the physical layer. 4
The hop-by-hop routing method is not adequate to cope with
wireless systems because every node of the wireless network must store the
location of every mobile system to which a route exists, which represents a
large amount of information. It is
impossible to keep routing information up to date and consistent throughout the
network, since a very large number of mobiles will exist. A possible solution would be to have each
mobile controller node with network-layer software enhanced by an additional
sublayer that performs routing to mobile systems. 4
Another major drawback is the fact that establishing virtual
connection takes longer. In a wide or
local area network, virtual connection establishment is fast, but it is likely
to take longer time in wireless networks.
It is not practical to re-establish all virtual circuits whenever a
mobile moves between pico-cells. A
possible solution would be to isolate small-scale mobility of the mobile from
the rest of the wired network. 4
Wireless ATM also embodies high noise interference and poor
physical level characteristics. Since
wireless ATM is designed for low-level error control, it cannot deal with
large-scale networks. Wireless ATM may
use either 16 or 24 byte cells, so segmentation and reassembly is required,
which allows one to use a segment counter that uses the two least significant
bits of the error control sequence number to try to cope with errors in these
networks. Another disadvantage is
finding a suitable channel sharing media access control technique at the data
link layer. Shared media access leads
to poor quantitative performance in wireless networks. 4
The Magic WAND (Wireless ATM Network
Demonstrator) was a joint European project established to specify and implement
a wireless access system for ATM. The
project started in 1996 and was projected to end in 1998. Large corporations such as Nokia, Intracom,
IBM, Lucent Technologies, and Eurecom, sponsored it. 5
The major goals of the project were
to specify and implement an access system for ATM networks that maintain the
service characteristics and benefits of the ATM to the mobile users. It also promoted the standardization of
wireless ATM access. The last objective
set by this project was to demonstrate that wireless access to ATM was
technically feasible, thus capable of providing real-time multi-media services
to mobile users. This would have been
demonstrated through user trials in hospitals where doctors would access
databases wirelessly, retrieve patient information from the network, consult
expert doctors, and share documents.
The user trials would also be conducted in universities where college students
would “surf the web” at high speeds.
5
The WAND project was to consist of
three major phases. In the first phase,
system and component design would be undertaken. In the second phase, the implementation of the design would be
started. In the implementation, the
mobile terminals would be as close as possible to the access points. In the final phase, the wireless ATM network
would be tested with real life users to verify the correct operation. 5
Although the project was never
completed, some accomplishments did occur. The first achievement was the
finalization of specifications by the project group for wireless ATM. It was decided that the communication
between the mobile units and the access points was to be set in the 5 GHz
frequency range. The WAND radio was to operate in the 5.15-5.3 GHz frequency
bands. The maximum theoretical
bandwidth was set at 20 Mbps, the theoretical Bit Error Rate (BER) at 10-6,
and the range of communication between the mobile units and the access points
was set up to 50 meters. 5
The project was productive in the
area of standardization. The project
designers wanted the Magic Wand to set the standard for future research and
study in the area of wireless ATM. The
radio channel model was developed and verified by measurements in the 5 and 17
GHz frequency bands. However, since no
further work was completed by the Magic Wand project its research was
terminated in 1997 and wireless ATM was never implemented. 5
Since
wireless ATM was never fully implemented in real life applications, the options
for the simulation of the behavior in the network are limited. Therefore, our simulation will focus on the
physical layer and examine the effects of noise and mobility on a wireless
channel. In the first Matlab program,
we analyze the detrimental effects of random noise on the availability of a
channel. Making the assumption that
noise is of random nature, we generate a random noise matrix with values
ranging from zero to one. In the
matrix, we assume that values below 0.4 indicate a channel that is not available
for transmission. Knowing that wireless
ATM uses OFDM (Orthogonal Frequency Division Multiplexing) with 16 channels, we
set the noise matrix to be 16x1. Running the program through a user specified
number of trials, the user is able to calculate the probability of finding a
good channel. The plots generated
represent the number and probability of having useful channels present at any
given time. Once all probabilities are
calculated for the specified number of trials, the average probability is displayed.
The
second Matlab program examines the added effect of mobility on
transmission. Continuing with the work
from the first program, we introduce a mobility factor in determining whether a
channel is suitable for transmission.
The program runs through a fixed number of trials (3) where the user
specifies the number of movements that occur for each trial. The movements are controlled by a random
matrix, where the values are scaled to indicate user mobility. Values below 0.5 are scaled using the formula
4*(x-1) to indicate movement towards the source, and values above 0.5 are
scaled using the formula 4*x to show movement away from the source. The movement values are then used to alter
the noise matrix. The assumption used is
that at 50 meters the signal becomes useless with no noise. Therefore, it was determined that the signal
will change by 1.4% per meter moved.
This value determines the amount of change a movement causes in the
noise matrix. When the simulation is
run, it generates three pairs of plots.
Each pair signifies both a plot of the movements, based upon a beginning
point of 25 meters away from the source, and a plot of the number of good channels
after each movement.
The
results of the simulations showed dire conditions for the use of ATM in
wireless networks. The first simulation
suggests that at any given time, there are only about 60% of the channels
available for transmission. While 60%
seems like a reasonable percentage, when compared to the almost 100% usability
of fiber optic cable that wired ATM achieves, it is severely lacking. To work dependant on having 60% of the
transmission channels available, wireless ATM will require incredible tweaking
of the wired ATM protocol.
The
second simulation adds on to the results of the first implying that mobility
will cause an even greater problem for the transition to wireless ATM. The simulation shows that moving even
slightly further away from the source can cause the number of available
channels to drop significantly. While
on the other hand, moving closer causes a greater number of channels to become
useful until the point where all 16 wireless channels can be active.
Our simulations, while only of the physical layer, show that much work needs to be done before ATM can become wireless. The effects of noise and mobility on the wireless channel create a starkly distinct environment from that of an error-free fiber optic cable. For these reasons and others, current research has now shifted to wireless IP-based systems. This shift is mostly due to IP’s built-in ability to handle transmission errors in the realm of wireless communications.
1)
Ayanoglu, E.,
K.Y. Eng, M.J. Karol, “Wireless
ATM: Limits, Challenges, and Proposals”, http://citeseer.nj.nec.com/cache/papers2/cs/1647/http:zSzzSzwww.exit109.comzSz~enderzSzwatm.pdf/wireless-atm-limits-challenges.pdf,
1996.
2)
Black, U. “ATM: Foundation for Broadband Networks”,
Prentice Hall, 1995.
3)
Peterson, L. and B. Davie.
Computer Networks: A Systems Approach. 2nd edition, Morgan Kaufman, 2000.
4)
Wasi, Atif S., “Wireless ATM”, ftp.netlab.ohio-stat,edu/pub/es/cis788-95/wireless_atm/index.htm,
1995.
5)
“Wand Overview”, www.tik.ee.ethz.ch/~wand/SUMMARY/WAND_97.htm,
1996, 1997, 1998.