7
Voice and Data
Convergence
7.1 INTRODUCTION
The subject of this book is all about the convergence of voice and data
services into a new world of advanced (and hopefully useful) applications
that will enrich the way people work and communicate. So a specific
chapter entitled voice and data convergence could jar a little with the
reader, however, this chapter was prompted by the words in the Interna-
tional Telecommunications Union telecommunications (ITU-T) I.121
specification ‘‘The need to integrate both circuit- and packet-transfer
mode into one universal broadband network’’. It is interesting to reflect
that what we now start to see becoming achievable through Internet
Protocol (IP) was stated as a goal in an ITU-T specification in the late
1980s.
One of the difficulties in finding a place for Asynchronous Transfer
Mode (ATM) in the first part of the book was that ATM is a packet tech-
nology, but has a circuit switching heritage, so if the reader will forgive
me, I have chosen to pick up ATM under a chapter all of its own under
voice and data convergence.
The simple statement (amongst others) in I.121, gave rise to the speci-
fications for Broadband ISDN (B-ISDN) and through the work of the ATM
forum the specification of asynchronous transfer mode. This chapter
explores ATM as a key enabler to convergent communications and appli-
cations. ATM has suffered against transport control protocol/Internet
protocol (TCP/IP), as TCP/IP started getting all the attention and ATM
seems to have been backwatered as a transmission mechanism.
That’s not so say ATM isn’t a success, quite the contrary, ATM has had a
Next Generation Network Services
Neill Wilkinson
Copyright q 2002 John Wiley & Sons, Ltd
ISBNs: 0-471-48667-1 (Hardback); 0-470-84603-8 (Electronic)
resounding success and in fact has found its way into public data
networks in a big way and may continue to gain success as a mechanism
for creating connection-oriented paths through a network that carries
Multi Protocol Label Switching (MPLS) (tag-) routed IP packets. MPLS
and ATM share a common goal, the idea of being able to relay packets at
high speed with quality of service. Some might argue that this is a waste
and in fact, in the introduction on packet technologies I state that IP,
combined with MPLS routers, could be carried directly over Dense
Wave Division Multiplexing (DWDM). The truth of it is that ATM,
MPLS and IP will be intimately bound together over the next few years
and maybe as technologies for creating even higher speed MPLS routers
emerge, then ATM will get left behind?
7.2 ASYNCHRONOUS TRANSFER MODE
Asynchronous transfer mode is a packet switched (or more accurately cell
switched) technology that has evolved from a circuit switched world! This
inheritance means ATM is a connection-oriented technology designed for
use in the backbone of carrier networks. Its Integrated Services Digital
Network (ISDN) and Signalling System number 7 (SS#7) heritage means
that the standards define two different types of protocol for connection
set-up and teardown. An interior protocol called broadband ISDN user
part (B-ISUP, cousin to the SS#7 call connection protocol ISUP) for the
connection between ATM switches (called the Network Node Interface
(NNI)) and an exterior protocol Q.2931 for the outside edge of the
network (called the User to Network Interface (UNI)). The standards
also go much further than this and go to great length to build in manage-
ability, again based on its heritage without the management aspect ATM
wouldn’t have got anywhere in a telco environment and of course quality
of service capabilities.
We start our look at ATM with the B-ISDN reference model, because
this is where it all started. The B-ISDN reference model is based on the
Open Systems Interconnection (OSI) reference model (remember we
mentioned this in the section on SS#7, Chapter 1. It’s incredible where it
just keeps turning up) and the original work on the ISDN standards. The
B-ISDN model contains three planes (columns if you like): the User plane
(U-plane), the Control plane (C-plane) and the Management plane (M-
plane) (Figure 7.1).
The C-plane is where connection control resides; it is here that you will
find Q.2931 and B-ISDN for control of connection set-up and teardown.
The B-ISDN model supports both a Switch Virtual Circuit (SVC) and a
Permanent Virtual Circuit (PVC) connection mode. The C-plane is respon-
sible for the management of the SVCs. PVCs are constructed via main-
tenance action and are thus the responsibility of the M-plane.
VOICE AND DATA CONVERGENCE94
The M-plane is responsible for operations, administration and mainte-
nance. The M-plane also has responsibility across the other planes and
between the planes to ensure management function can monitor and
control all the aspects of the B-ISDN stack.
The U-plane is simply the place where application level protocols and
functions reside. In this context TCP/IP is counted as an application level
protocol to the B-ISDN service. Other examples of elements/services that
reside in this plane are video on demand and Voice over IP (VoIP).
Sitting below each of these plane functions are two additional layers,
the ATM layer and the physical layer. Between the ATM layer and the U-,
M- and C-planes is a glue function called an adaptation layer that allows
these planes to utilise the cell-based structure of ATM.
The Signalling ATM Adaptation Layer (SAAL) glues the C-plane to the
ATM layer. The ATM Adaptation Layer (AAL) functions for the C-plane
are further segregated into AALs based on the type of service/traffic the
connection (SVC or PVC) will carry (this will be expanded on later). The
M-plane also uses an adaptation layer to carry management information
and control protocols between the ATM nodes in the network. Examples
of these protocols are the Simple Network Management Protocol
(SNMP),
1
Common Management Information Protocol (CMIP)
2
and
Local Management Interface (LMI).
3
There is a subtlety in the use of the AAL, for the U-plane, it only
operates at the edge of the network not in the core, for the C- and M-
planes the AAL is used both at the edge and in the core. For example
SAAL is used to carry UNI signalling (Q.2931) at the edge and to carry
NNI signalling (B-ISUP) in the core.
How are different media with different loss, delay and bandwidth
requirements carried across a fixed cell size network (we’ll come to the
fixed cell size later)? This is the role of the adaptation layer. As we have
already seen voice has a number of constraints with regards to delay and
jitter, the same is true for high quality video-on-demand services (some of
the video-on-demand requirements can be overcome with other techni-
ques). Data services such as email and web browsing are less demanding
of the network, however, loss of data for these services is a problem,
whereas loss of the odd voice or video sample is less of an issue. The
AAL has to deal with these differences and be able to place all of these
different requirements into the ATM fixed 53-byte cell structure.
The ATM standards organisation decided that the ATM adaptation
layer be split into two sublayers: the Convergence Sublayer (CS) and
7.2 ASYNCHRONOUS TRANSFER MODE
1
An Internet standard for exchanging management information between network nodes
and a network manager.
2
ITU-T standard for exchanging management information between network nodes and a
network manager.
3
Used in frame relay networks for exchanging status information between devices such as
routers.
95
the Segmentation And Reassembly sublayer (SAR). As the name suggests
the SAR sublayer is responsible for breaking the application level payload
down into 48-byte chunks to fit in the cell. The application level data may
be anything from 1 byte to many thousands of bytes (remember an IP
packet can theoretically be in excess of 65,000 bytes in size). The ATM
payload may not carry all application level information there may also be
some adaptation layer headers and trailers. A series of classes of traffic
have been defined by the standards (classes A–D and X), each class repre-
sents the characteristics of the application traffic it represents: whether it is
synchronous or asynchronous, whether it is a constant bit rate source
(voice) or a variable bit rate source and whether the service is connection
or connectionless.
In order to support these different types of traffic, a different ATM
Adaptation Layer (AAL) is defined to provide a service to each of these
classes (AAL1–5, with AAL3 and 4 combined during a revision of the
standards). AAL1 can be used to deliver 64 kbps voice through a constant
bit rate service; AAL2 provides a variable bit rate service for example for
compressed voice with silence detection (e.g. G.729a). AAL3 and 4 for
connectionless data services and AAL5 is a refinement of the AAL3 and
4 specifications.
Why a 53-byte cell size and a 48-byte payload? The reasoning on size of
cells was one of compromise between the data people who wanted a large
cell size to carry more information in each transmission unit (let’s face it
the line rate is going to be many megabits per second, so it wouldn’t take
long to transmit a large cell and if you lost one, not a problem just retrans-
mit it), and the voice people who wanted a small cell size. The voice
people were looking to small cells to minimise latency in buffers on
VOICE AND DATA CONVERGENCE96
Figure 7.1 B-ISDN model
ports on the ATM switches and looking at reducing any problems asso-
ciated with packet loss (the smaller the frame, the fewer voice samples per
frame, the fewer samples lost if a frame gets dropped).
All of these cells then need to be multiplexed on to the appropriate
physical layer (Synchronous Digital Hierarchy (SDH), Synchronous Opti-
cal NETwork (SONET), DWDM or even digital subscriber lines for exam-
ple). This is the role of a transmission convergence (TC) sublayer.
We stated earlier that ATM is a connection-oriented service, so how are
connections managed and for that matter created? The 5-byte header field
on each ATM cell contains identifiers: Virtual Path Identifier (VPI) and
Virtual Channel Identifier (VCI), the two combined representing a virtual
circuit identifier and a virtual path is a collection of virtual channels going
between two points. These fields represent the logical connections that the
data in the payload (48 bytes) are being transported on. The connections
can be pre-provisioned connections (Permanent Virtual Circuits (PVCs))
or on-demand connections (Switched Virtual Circuits (SVCs), although
this term is out of favour). The important part of this nomenclature is the
word virtual. Whilst a connection context exists using a particular combi-
nation of VPI and VCI fields, if the endpoint of the connection is not
generation data, no cells are transmitted for this endpoint, unlike Time
Division Multiplex (TDM) for example where, even if there is no speech
timeslots are transmitted across the network.
The important point about the use of the VPI and VCI fields is that their
significance is local to the switch node they are passing through, they do
not have global significance unlike for example an IP address. Clearly
globally significant VPI and VCI addresses wouldn’t be possible; there
just aren’t enough bits in the ATM cell header. So this begs the question
how is a connection across multiple switch nodes in an ATM system
created? Figure 7.2 shows an example of the use of VPI and VCI fields.
In fact the usage of VPI and VCI is not defined by the ATM standards and
is left to the implementation.
ATM’s Quality of Service (QoS) aspects include monitoring and control
of such aspects as cell loss ratio, cell delay, and delay variance. QoS in
ATM is defined as an end-to-end characteristic, the properties such as cell
loss ratio are a measure based on this fact. The ATM forum, in a measure
to simplify QoS for users, has defined five classes to express the service
characteristics at the UNI. These are:
1 Class 0: No QoS guarantees, best effort service
2 Class 1: Constant Bit Rate applications (CBR) for example circuit
emulation
3 Class 2: Variable Bit Rate (VBR) real-time traffic for example
compressed packetised speech
4 Class 3: Connection oriented services
5 Class 4: Connectionless protocols
7.2 ASYNCHRONOUS TRANSFER MODE 97
It is left to the network operator to assign the values of cell loss ratio, cell
delay, etc. to each of these classes and how best to manage these factors in
their network. QoS is a complex topic and requires more consideration
than is available in this text.
Clearly there is a lot more to both broadband ISDN and ATM than this
short chapter covers, it is hoped this chapter gives a flavour of the power
of ATM and its use for a multimedia carrier. Significantly more detail is
covered in [BLACK2, DYSA] and are a recommended read if you want to
(or need to) know more.
VOICE AND DATA CONVERGENCE98
Figure 7.2 ATM VPI and VCI usage