Escolar Documentos
Profissional Documentos
Cultura Documentos
Communications
(Part I)
OBJECTIVE
On completion of this topic, you should be able to:
1.
2.
3.
4.
5.
6.
Describe the ATM cell header structure and the virtual path
concept.
7.
8.
9.
X OVERVIEW
This topic describes the general principles and structures of two advanced high
speed backbone network technologies, and gives you the chance to start
evaluating their potential. Well start our discussion by defining what we mean
by high speed networks, and then well introduce the two advanced high speed
networks. The first, Gigabit Ethernet, is an upgraded version of the 802.3
22 X
INTRODUCTION
We begin our discussion here in this topic with high speed backbone networks,
the infrastructure to support the growing bandwidth-greedy applications.
First, we need to ask an obvious but important question: what are high speed
networks? Well, if you think about it, you will agree that the answer changes
with time. Recall that modems in the 1980s were still running at speeds of
300~1200 bps, and networks (in the early 1980s) carrying a data rate of around 10
Mbps were considered to be high speed. In fact, networks such as Ethernet
could easily support hundreds of users in a local area with applications such as
email and file transfer.
23
The introduction of the World Wide Web (WWW) in the 1990s, however, caused
explosive growth in data traffic over the Internet. The old high speed networks
were no longer sufficient to support the continuous growth in traffic volume. In
addition, there were increasing demands to support multimedia applications
over long distances; for example, to set up a video conference between two
parties in the US and Hong Kong. To support these new types of services (such
as video), advanced compression techniques to pack the data but also new
high speed networks to transfer the information needed to be developed.
Table 2.1 below tracks how the meaning of high speed networks has evolved
over time. This is similar to the evolution of computers, in which faster and faster
CPUs like 8086, 80286, and so on, have been developed until today, when the
Pentium III has captured the market.
Table 2.1: The meaning of high speed networks changes with the times
Year
1960s
64 kbps
1970s
1 Mbps
1980s
10 Mbps
Early 1990s
100 Mbps
Today
1 Gbps
2005
> 10 Gbps
24 X
25
26 X
READING 2.1
Moorthy, V (1997) Survey of Gigabit Ethernet, University of Ohio
<http://www.cis.ohio-state.edu/~jain/cis788-97/gigabit_ethernet />
Although this paper was written before the approval of the current
standard (1997), it provides a solid background on the development
and basic concept of Gigabit Ethernet.
READING 2.2
Introduction to Gigabit Ethernet, Technical Document from Cisco
Systems
http://www.cisco.com/warp/public/cc/techno/media/
lan/gig/tech/ gigbt_tc.htm
This brief article provides a technical overview of Gigabit Ethernet,
including its architecture, standards and topologies.
27
READING 2.3
White paper: Gigabit Ethernet Overview updated May 1999, Gigabit
Ethernet Alliance http://www.10gea.org/Tech-whitepapers .htm
The Gigabit Ethernet Alliance is an open forum whose purpose is to
promote industry cooperation in the development of Gigabit Ethernet.
This white paper provides a brief overview of Gigabit Ethernet,
including the advantages of Gigabit Ethernet, Gigabit Ethernet
technology fundamentals and network migration scenarios.
As Gigabit Ethernet is still under development, you are encouraged to
visit the Gigabit Ethernet Alliance website for updated information.
There are also a couple of optional readings you might want to look
up if you want more details or are particularly interested in this
subject:
(Optional) FAQ about Gigabit Ethernet, Gigabit Ethernet Alliance,
http://www.10gea.org/tech-faqs.htm
This website contains common FAQ about Gigabit Ethernet. Interested
readers can refer to them for further information.
(Optional) Philippe Ginier-Gillet and Christopher T Di Minico,
After the above readings, it is time to test your understanding of the material.
Please do your best to answer the following questions.
28 X
ACTIVITY 2.1
1.
2.
3.
4.
2.
29
30 X
31
2.
3.
The aims of ATM were too aggressive and it tried to support too many
services. This slowed down the process of defining standards.
4.
32 X
5.
The explosive popularity of WWW made the Internet the de facto global
network. This made it harder to persuade customers to change to ATM.
33
only the virtual circuit needs to be identified, but not the exact address of the
destination. Also, with the help of a fixed route between source and destination,
the packets can arrive in the same sequence as they were sent from the senders
side.
Please read the following reading to refresh your memory of different switching
technologies circuit switching, packet switching and virtual circuit switching.
READING 2.4
Walrand and Varaiya (1996) High-performance Communication
Networks, Morgan Kaufmann: Section 2.5.2, pp. 6065.
This reading reviews the concepts of circuit switching, packet
switching and virtual circuit switching.
34 X
Suppose the restriction of dedicated time slots is removed, however, and other
sources are allowed to transmit as long as their buffers are non-empty. In Figure
2.1(b), the second packet from source 3 is allowed to transmit after the second
packet from source 1. Then no time slot is wasted, resulting in higher utilization
of the link capacity. In fact, there are no empty slots as long as the sources have
packets to transmit. This is known as statistical multiplexing (or asynchronous
time division multiplexing) because the slots are no longer used synchronously.
This is the reason why the term asynchronous is used in ATM. However, each
packet must carry an address to identify itself because its identity can no longer
be inferred by its time slot position in a frame.
Figure 2.1(b): All the time slots are used in statistical multiplexing
35
would like to allocate sufficient bandwidth (less than the sum of peak rates) to
support multiplexed traffic without losses.
In Figure 2.2, you can see three traffic sources multiplexing into an output link.
The sum of their peak rates and the sum of their average rates are 5 Mb/s and 2.5
Mb/s. To ensure no packet losses, we can allocate a bandwidth of 5 Mb/s to the
output link. However, since not all three traffic sources transmit at their peak bit
rates simultaneously, a link capacity of 3 Mb/s is enough to serve the
multiplexed sources. That is, we have a saving of (5 3) = 2 Mb/s.
The ratio of total input capacity to the total output capacity is defined as the
multiplexing gain. In this example, the multiplexing gain is (5/3). In general, the
multiplexing gain depends on the number of traffic sources and traffic
characteristics. For ideal cases, we would like to have the link capacity equal to
the sum of average rates of multiplexed traffic.
On the other hand, weve lost some simplicity of implementation by using
statistical multiplexing. We need to have additional overhead such as packet
labels to identify source or destination addresses. This requires more complicated
network controls to ensure the correct routing of packets. In addition, the actual
savings in bandwidth are highly dependent on the traffic characteristics. If the
traffic is very smooth, we will not see many savings in bandwidth. Therefore, we
have a trade-off between complexity and multiplexing gains.
In any case, you can see now how statistical multiplexing is used in ATM to
transfer multiple traffic streams over the networks. The following sections
describe some more of the basic concepts behind ATM.
36 X
SELF-TEST 2.1
What are the advantages and disadvantages of using statistical
multiplexing over time division multiplexing?
Benefits of ATM
Several benefits of ATM are described in the following reading. Please go
through them carefully.
READING 2.5
The International Engineering Consortium, Asynchronous Transfer
Mode (ATM) fundamentals, Web ProForum Tutorial, Section 2, pp.
14.
<http://www.webproforum.com/atm_fund/>
SELF-TEST 2.2
List and explain the advantages of using ATM over existing LANs.
37
2.
3.
4.
For each plane, separate layers perform specific functions. Lets look at the lowest
three layers: the physical layer, which mainly transports information (bits/cells);
the ATM layer, which mainly performs switching/routing and multiplexing; and
the ATM adaptation layer (AAL), which adapts service information to the ATM
stream.
The physical layer is responsible for the correct transmission and reception of bits
on the physical medium. Common physical media such as UTP, coaxial cable
and optical fibre can support ATM.
The ATM layer is fully independent of the physical layer. ATM provides a
connection-oriented cell switching service using the VP concept. The ATM layer
is primarily responsible for the generation of the cell header and the functions
associated with the header, such as the switching and routing of cells, flow
control, congestion notification, bit error detection in the header, etc. Some of the
main functions of ATM headers will be explained in the next section.
The ATM Adaptation Layer (AAL) is an end-to-end protocol that provides the
interface between the ATM Layer and higher layer protocols and applications. It
is the capabilities of the AAL that make ATM able to support a wide range of
services. We shall discuss AAL in a later part of this topic.
38 X
39
READING 2.6a
Tanenbaum, A (1996) Computer Networks, Prentice Hall: Section
5.6.1, pp. 450452.
This reading examines the ATM cell structure and explains the
function of different fields in the ATM header.
SELF-TEST 2.3
What are the disadvantages of using a 5-byte header in an ATM cell?
40 X
get new seat numbers (a new VCI) in Tokyo, while some passengers will stay in
their seats on flight UA001 to continue their trip to Los Angeles. When the plane
arrives in Los Angeles, all passengers leave the plane (VP ends). The flight
number of the plane may change to UA311 (a new VP) and start a journey to
Boston.
VPI/VCI addresses are only meaningful to local connections within the VC or
VP. Thus there is no need for global address as there is in TCP/IP. This can
greatly reduce the size of VPI/VCI addresses. To extend our airline analogy, the
flight number and seat numbers are only meaningful to one particular airline
(United Airlines in this case).
This virtual path concept is adopted in the ATM standard. The VPI provides an
explicit path identification for a cell, while VCI provides explicit circuit
identification for a cell. Basically, a virtual path is a bundle of virtual circuits that
are switched as a unit by defining one additional layer of multiplexing on a percell basis underneath the VCI. A predefined route is provided with each virtual
path; thus it is not necessary to rewrite the routing table at call set-up. Therefore,
call-by-call processing at switching nodes is reduced, and call set-up delay is
decreased. That is, VPs are semi-permanent connections. To take full advantage
of the VP concept, each VP is allocated a fixed amount of bandwidth. To establish
a connection between two end-stations, the networks simply find a suitable path
consisting of a sequence of VPs that has sufficient bandwidth to support the
connection. In other words, the VP concept provides a logical link between two
nodes, and this creates an overlay network with logical links on top of the
physical network. Now why not re-read this paragraph and think of it in terms of
the UAL flight to Los Angeles?
In short, the VP/VC scheme gives lower node cost but at the expense of higher
link cost. This characteristic is valuable to ATM networks because large
bandwidth will be available as high capacity optical fibres become more widely
used. A transmission path may contain several virtual paths and each virtual
path may carry several virtual circuits.
The next reading examines the process of connection set-up and routing in ATM
using the VP concept.
READING 2.6b
Tanenbaum, A (1996) Computer Networks, Prentice Hall: Sections
5.6.25.6.3, pp. 452458.
41
SELF-TEST 2.4
What are the advantages of using the virtual path concept in terms of
network management?
2.
3.
4.
42 X
The ITU-T has originally recommended AAL types 3 and 4 for connectionoriented and connectionless services, respectively. Since these two supported
services are almost the same, the ITU-T has combined them to form the AAL
3/4. AAL3/4 packets are used to carry computer data.
Since the AAL type 3/4 was originally proposed to support connectionoriented services as well, it has a high overhead. As a result, AAL type 5 is
recommended to offer services with less overhead and better error detection.
The ITU-T has adopted AAL type 5 for IP traffic and for signalling as well.
AAL type 5 has been shown to have a 2030% better throughput than that of
AAL type 3/4.
on the type of service, different quality of service (QoS) thresholds must therefore
be specified.
When the network accepts a connection request, the user and the network agree
on a traffic contract for the duration of the connection. With this contract, the
network guarantees the requested service demand of the connection as long as
the source traffic complies with some specified limits. Accordingly, a traffic
contract includes traffic descriptors, service requirements and the conformance
definitions. Some of the parameters adopted by ATM Forum include:
1.
Peak Cell Rate (PCR) the inverse of the minimum time between two cell
submissions to the network, i.e., a connections maximum transmission
rate;
2.
3.
Minimum Cell Rate (MCR) the inverse of the maximum time between
two cells; that is, the minimum usable bandwidth for a connection;
43
4.
5.
These parameters are the traffic descriptors to specify the behaviour of the traffic,
and they are defined by the Generalized Cell Rate Algorithm (GCRA), which
controls the arrival times of cells.
There are additional parameters describing the service requirement:
1.
Cell Loss Ratio (CLR) the ratio of the number of lost cells to the total
number of cells sent by a user over the lifetime of the connection;
2.
Cell Transfer Delay (CTD) the delay experienced by a cell inside the
network including coding and packetization delay, propagation delay,
transmission and switching delay, queuing delay, and re-assembly delay,
etc.;
3.
Please note that not all the QoS requirement parameters and traffic descriptors
are used to specify the service class. Different service classes may specify only
some of the above parameters.
2.
3.
4.
5.
6.
44 X
Please read Sections 2.12.3 of Traffic Management Specification Ver. 4.1 from the
ATM Forum for a detailed explanation of each service class. Pay particular
attention to Table 2.1 for the relationship between the service classes and their
corresponding traffic and QoS parameters. Some examples of different
applications for ATM service classes are also discussed in these assigned
readings.
READING 2.7
ATM Forum (March 1999) Traffic Management Specification Ver. 4.1,
Sections 2.12.3, 3.13.2, pp. 48, 12 <ftp://ftp.atmforum.com/
pub/approved-specs/af-tm-0121.000.pdf>
READING 2.8
ATM Forum (March 1999) Informative Appendix IV: Applications
Examples for ATM Service Categories, Traffic Management
Specification Ver. 4.1, and p. 95 <ftp://ftp.atmforum.com/pub/
approved-specs/af-tm-0121.000.pdf>
The very long document from the ATM Forum that the above two extracts are
taken from contains the most updated specifications for ATM traffic
management. Although the document is very comprehensive, it is too detailed
for the purpose of this course, so you should only read the sections specified
here.
READING 2.6c
Tanenbaum, A (1996) Computer Networks, Prentice Hall: Sections
5.6.45.6.5, pp. 458463.
(O
Optional) The International Engineering Consortium, Asynchronous
Transfer Mode (ATM) Fundamentals, Web ProForum Tutorials: Section
4, pp. 67 <http://www.webproforum.com/atm_fund/>
45
SELF-TEST 2.5
Give an example of an application in each service class.
46 X
schemes are less desirable than simple schemes that can be easily implemented in
high speed hardware. Therefore, a preventive control method has been proposed
to prevent congestion by taking appropriate action before the congestion actually
occurs.
Nevertheless, it is generally agreed that preventive control techniques are not
sufficient to eliminate congestion problems in ATM networks, and that when
congestion happens, it is necessary to react to the congested state. In addition,
preventive controls may be inefficient because bandwidth may be over-allocated.
Reactive control techniques are used to initiate recovery from a congested state
so that the network will not be stuck with long-term congestion.
Accordingly, the congestion control schemes in ATM can be further classified
into different layers according to different time scales. This multilevel congestion
evaluation and control is motivated by the fact that communication terminals
often have traffic states characterized by these levels. Thus, to improve network
efficiency, traffic control should be exercised at these levels.
The first of the following two readings presents an overview of the traffic
management issues and possible solutions to support multimedia over an ATM
network. You will find that different levels of management mechanisms are used
to control the traffic. The next reading, which is optional, explains the different
levels of management and control in ATM networks.
READING 2.9
Zheng, B and Atiquzzaman, M (1999) Traffic management of
multimedia over ATM networks, IEEE Communication Magazine,
37(1) (Jan.): 3338.
(Optional)
The
International
Engineering
Consortium,
Asynchronous Transfer Mode (ATM) Fundamentals, Web
ProForum
Tutorials,
Section
9,
pp.
1415
<http://
www.webproforum.com/atm_fund/>
ACTIVITY 2.2
Please briefly describe three different traffic management strategies
in ATM networks.
47
2.
Traffic shaping
One major advantage of ATM is the statistical multiplexing gain of bursty traffic.
Now, suppose we have a source sending 100 Mbps for 1 second and remaining
silent for 99 seconds. Thus, the average rate of the source is only 1 Mbps over a
period of 100 seconds. Since 100 Mbits of information can be dumped into the
network during the 1 second period, this highly bursty traffic source can
jeopardize the whole network. The situation is similar to sending thousands of
email bombs to an email server at the same time. As a result, a control
mechanism is introduced to force the traffic to be transmitted at a more
predictable rate. This preventive mechanism is known as traffic shaping.
Traffic shaping is to regulate the average transmission rate of data by smoothing
out bursty traffic. For instance, the smoothed traffic will only transmit 10 Mbps
over a period of 10 seconds, which can be easily handled by the networks. One
traffic shaping algorithm commonly used in ATM is the leaky bucket. In the
48 X
following reading, you will learn the basic concepts related to the leaky bucket
and token bucket algorithms.
READING 2.10
Tanenbaum, A (1996) Computer Networks, Prentice Hall: Section
5.3.3, pp. 379384.
SELF-CHECK 2.6
Please briefly describe two major differences between the leaky
bucket and the token bucket.
READING 2.11
Reading 2.11: White, P (1998) ATM switching and IP routing
integration: the next stage in Internet solution? IEEE
Communications Magazine, 36(4) (April): 7983.
49
READING 2.12
Reading 2.12: Dahod, A (1999) Viewpoint: ATM v. Internet Protocol,
IEEE Spectrum, 36(1) (Jan.): 3031 <http://www.spectrum.ieee.org/
spectrum/jan99/features/commsb1.html>
ACTIVITY 2.3
1.
2.
3.
50 X
READING 2.13
Rauch,
READING 2.14
Manchester, J et al. (1998) IP over SONET, IEEE Communications
Magazine, 36(5) (May): 136142.
51
READING 2.15
Azcorra, A et al. (1999) IP/ATM integrated services over broadband
access copper technologies, IEEE Communications Magazine, 37(5)
(May): 9096.
52 X
This topic discussed several high speed networks. We started with Gigabit
Ethernet, which is the extension of the popular 802.3 Ethernet network. Its
backward compatibility with 802.3 Ethernet is likely to ensure the quick
acceptance of Gigabit Ethernet in the LAN market.
We then focused our study on a new technology ATM, which is the
supporting technology for B-ISDN. ATM is a connection-oriented, packetswitching technique that uses short fixed-size packets called cells to transfer
information. The layered B-ISDN model for ATM was then briefly described, as
were the functions of different fields in the ATM header and the various types of
AAL to adapt the higher-layer services to services supported by the ATM layer.
ATM can support multi-type traffic with the QoS guarantee that is essential for
B-ISDN. Different traffic classes can be supported in ATM networks, and they
were specified in terms of traffic parameters. Traffic management is essential to
high speed networking. Various control mechanisms to support multimedia were
briefly discussed. We have also briefly introduced the basic concepts of call
admission control and traffic shaping.
The purpose of a computer network is interconnecting different nodes. We
described several approaches to support IP over ATM. We made a simple
comparison between Gigabit Ethernet and the ATM network. Then, we described
a competing technology IP over SONET.
Finally, we discussed the issue of providing ATM service to home users with
new xDSL technology. More information on the residential broadband services
will be described in other topic.
53
Asynchronous Transfer
Mode (ATM)
A
connection-oriented,
packet-switching
technique that uses short fixed-size packets called
cells.
Broadband Integrated
Service Digital Network
(B-ISDN)
Gigabit Ethernet
IP-over-ATM
Multiplexing gain
Statistical multiplexing
Traffic shaping
Virtual channel
54 X
Virtual path
Wireless ATM
xDSL
Different
digital
technologies.
subscriber
lines
across
When the traffic is bursty, statistical multiplexing is more bandwidthefficient. The multiplexing gain of statistical multiplexing can be much larger
than that of time-division multiplexing.
When the traffic is smooth (e.g. constant bit rate), time-division multiplexing
can be more bandwidth efficient.
Self-test 2.2
Advantages of ATM:
High performance via hardware switching the small header and fixed
packet size reduce the processing delay.
55
Self-test 2.3
Disadvantages for using 5-byte header in an ATM cell:
Large overhead because of the space in each cell used for ATM header. Thus
the maximum efficiency of bandwidth is only around 90%.
Self-test 2.4
Advantages of VP concept:
For instance, if a company wants 10 Mbps bandwidth, you just set up a VP with
10 Mbps bandwidth to this company. This is similar to hiring a circuit.
Self-test 2.5
CBR
RT-VBR
NRT-VBR
ABR
UBR
Telephone
traffic, CBR
video
Interactive
compressed
video
Multimedia
email
File transfer
TCP/IP traffic