Voice over Digital Subscriber Line (VoDSL) deployment is
well underway -- and although these are early days, research
firm TeleChoice, Inc.,
estimates that ILECs and CLECs will deploy more than one
million voice-enabled DSL lines by 2003. For those service
providers beginning down the VoDSL path, today's solutions
leverage existing DSL and PSTN infrastructure.
This article explores the current state of VoDSL and the
next steps in VoDSL evolution, including the current methods
of integration with Local End Offices (Class 5 switches),
early examples of Local End Offices replacement, and the
migration toward next-generation network architecture.
Market pressures have never been greater on traditional ISPs
and CLECs. Having enjoyed generous access to capital in
recent years -- and having invested massively on broadband
deployments -- most are not yet realizing adequate return on
investment -- and thus face financial distress. VoDSL can
provide a major boost to service providers' profitability,
because it is riding the wave of market trends. Despite
having reached an inflection point, probably in 2000, where
the amount of worldwide data traffic exceeded worldwide
voice traffic, voice still pays the bills. In fact,
according to the Gartner Group, voice will still account for
more than 75 percent of the total data and fixed telephony
market in 2003! Voice is the "killer application" -- and
bundling voice with data access is a critical service
provider survival strategy.
VoDSL provides all the capability of DSL -- and more.
Specifically, while DSL is restricted to carrying data
packets to and from a customer premise to the central office
over standard, existing copper wire (and then on to the
ISP), VoDSL allows that same bandwidth to support both voice
and data, on an as-needed basis. For example, where, with
DSL a single analog voice line can typically carry up to 1.5
Mbps of data traffic, with VoDSL that same 1.5 Mbps of
bandwidth can now support as many as 16 voice channels, as
well as substantial data traffic.
Importantly, this enhanced ability to support voice as
well as data requires no major infrastructure changes,
relying on the same basic architectural components as DSL --
a DSLAM (digital subscriber line access multiplexer) in the
Central Office (CO) as well as the local end office operated
by the incumbent or the challengers. Over time, DSLAM
vendors have added Quality of Service (QoS) parameters to
their equipment, not only to properly handle both voice and
data packets, but to provide multiple classes of services
for data traffic to enforce Service Level Agreements (SLA).
This was relatively easy to do because most DSLAMs are based
on ATM -- and ATM QoS is standards-based and
well-understood. To achieve the benefits of VoDSL, the only
requirements are a new Integrated Access Device (IAD), and a
VoDSL gateway for converting packetized voice to TDM
Note that the infrastructure between the IAD and the
VoDSL gateways is all based on ATM. The customer premise
equipment is a VoDSL IAD, containing all the intelligence
needed to multiplex voice and data traffic and to allocate
bandwidth accordingly, with voice always taking preference.
Typically, from 8-16 voices lines are connected to the IAD.
The PCs, typically on a 10Base T LAN, are connected through
a hub to the IAD. For the voice side of the IAD, it will
packetize the information using ATM AAL2 formats, as well as
provide functions for signalling, ringing, echo
cancellation, and voice compression. On the data side, the
IAD encapsulates the traffic using ATM AAL5 formats and
provides functions such as firewalling, routing, DHCP, etc.
Multiple DSL loops are aggregated into the DSLAM. At the
point where they exit the DSLAM, the voice and data ATM
packets are routed to a regional packet network and into the
cloud. There, network switches separate the streams, sending
the data to broadband remote access servers (B-RAS: also
called a subscriber management system) and the voice to a
VoDSL gateway, also called an access gateway.
The VoDSL gateway converts the packets coming from the
DSLAM to TDM so that they can be transferred over T1/E1
lines (using GR-303 in North America and V5.2 in most of the
rest of the world) and into existing CO switches. Note that
because the VoDSL gateway converts ATM voice packets to the
TDM format, it really is a VoATM gateway.
Today's VoDSL access gateways require these four basic
- ATM-to-TDM conversion;
- Echo cancellation with at least 64ms tails;
- Compression, typically ADPCM; and
- Trunking (T1 interfaces with GR-303 protocol or E1
interfaces running V5.2 protocol).
AAL2 and Loop Emulation Service
The recent advent of ATM Adaptation Layer 2 (AAL2), a set of
standards developed by the International Telecommunications
Union (ITU) and the ATM Forum, make it possible for
compressed voice to ride alongside data and multimedia
streams on an ATM infrastructure with all ATM's classic
benefits: Scalability, mature standards, and dexterous
handling of QoS.
In addition to the fact that most DSL networks are
ATM-based today, ATM AAL2 is the perfect candidate for VoDSL,
because it provides the following capabilities:
- Multiple users channels on a single ATM virtual
- Bandwidth efficient transmission of low-rate, short
and variable packets in delay sensitive applications.
- Support for both constant bit rate (CBR) and variable
bit rate (VBR) classes of service for network traffic
optimization. The VBR class, with statistical
multiplexing, enables the treatment of compressed voice,
silence detection/suppression, and idle channel removal.
- Variable length payload across multiple cells for
Loop Emulation Service (LES) is a specialized subset of
AAL2 based on ITU-T specification 1.366.2 "SSCS for
Narrowband Services over AAL2." In a nutshell, LES supports
derived POTS and ISDN services, specifying management of
AAL2 channels, signalling details (both channel-associate
signalling [CAS] and common channel signalling [CCS]), and
in-band management for the IAD. LES is now the industry
standard for open communications between IADs and VoDSL
A Step Toward the Future
Today, voice and data convergence has been achieved in the
access part of the network while maintaining service
transparency and leveraging all the goodness of the PSTN --
well developed customer care, robust management and billing,
as well as the existing infrastructure (both DSLAMs and
Local End Office switches).
The next wave of convergence is beginning to hit the
Class 5 switches, in the form of entry-level local exchange
switching solutions. An interesting example is the Broadband
Voice Platform (BVP). You can think of this as a collapsed
Class 5 switch with a VoDSL interface. The BVP provides a
subset of Class 5 switch functionality, including SS7
signalling. It is not targeted at replacing the room-size
Class 5 switch, nor supporting all the Class 5
functionality, but is an alternative for more focused
deployments or simply for operators that may not want to
incur the capital expenses associated with a big Class 5
Over and above the classical VoDSL gateways, the BVP is
required to support tone detection and generation, as well
as the ability to support SS7 signalling into the PSTN.
Now, let's consider convergence at the local end office --
and the evolution toward end-to-end packet voice.
The Next Generation Network
Worldwide markets are driving the evolution to packet voice --
and it is the impetus for major investment. Whether the
ultimate goal of end-to-end packet voice will be met by ATM
or IP or both remains to be seen, but at this stage ATM is
superior to IP in its handling of end-to-end Quality of
Service -- and thus will be a key player in this migration.
Regardless of the end state, with as many as 2 billion
circuit switched (wireline and wireless) phones worldwide by
2004 (Coleago Consulting
Ltd. 2000) we will be working with hybrid networks for
many years to come.
To address the requirements of hybrid networks, one might
consider building a super switch to provide all the
functions of a Class 5 and handle packet voice and continue
to control all call processing for the myriad calling
features that exist today. However, this approach would
recreate the limitations of today's Class 5 switch: Lack of
scalability, inflexibility, high costs, and perhaps most
importantly, prohibitively long time to market for new
A fundamental aspect of the next generation network
architecture is the softswitch architecture. It consists in
taking the functionality of the classical Central Office
Switch (call control, signalling, feature control, etc.) and
exploding it into various platforms that perform more
dedicated functions. The key here is that decomposition
provides open interfaces between these essential network
functions. This evolution can be compared to the evolution
of the computing environment: Much like the movement from
mainframes and dumb terminals to today's client-server
architectures, the softswitch approach decentralizes the
control of the network, separating the call processing
function from physical switching function, while connecting
these via a standard protocol, mostly either a media gateway
control protocol (MGCP) or media gateway control (Megaco,
ITU standard H.248). Megaco is presently the culmination of
a plethora of standards that have emerged in the recent
years. Megaco is presently being standardized in the ITU
The key logical elements of the decomposed architecture
- Media gateway, which performs the conversion of one
medium (POTS lines, cable, DSL, 3G Wireless) onto
another packet-based medium (ATM or IP or both).
- Media gateway controller, (also known as a
softswitch), which performs the call processing
- Signalling gateway, which provides "translation" of
signalling messages between the PSTN and the packet
network. In an ATM network, this may take the form of
SS7 to Q.BICC protocol conversion. In an IP-based
network, this may be SS7 to SIP.
The typical switching fabric function that is at the
heart of any TDM switch becomes the packet network cloud,
which is populated with switches directing packets towards
their destination. The typical end office functionality then
becomes distributed and virtual, providing carriers a lot
more deployment agility.
The media gateway is the switching/bearer transport
platform -- it is the hardware on the edge of the network
that typically converts packet-to-TDM conversion under the
control of a softswitch. Media gateways (MG) will be the
junctions that provide a path between switched and packet
networks for such media as voice and fax. Media gateways can
either reside at the very edges of the public network at the
customer premises or in a central office.
The media gateway controller is the name of the
softswitch in the Megaco scheme. Media gateway controllers (MGC)
will serve as the brains of the operation because they will
control many media gateways.
The typical softswitch device provides call control and
routing functions for network voice and data traffic. It
also serves as a platform for services-creation capability
and is scalable. It also supports multiple applications and
SS7 capability. It is not embedded with the media gateway,
but stands alone, typically in a generic platform like a Sun
The signalling gateway provides the linkage between the
PSTN and the signalling methodology in the decomposed
architecture. Today, this takes many forms, but the work is
encompassed in the IETF Sigtran workgroup. Another
component, a signalling gateway, will play an analogous role
in capturing requests for address and enhanced service
information by enabling retrieval via IP from existing
SS7-enabled networks and their databases of
subscriber-related information. In turn, the relevant
address information is passed from a signalling gateway to
Like in client server architectures, there is not a
1-to-1 mapping between the softswitch's logical functions
and the network topology. In particular, the media gateway
control functions are distributed to provide scalability and
Megaco/H.248 is an emerging standard that will enable
voice, fax and multimedia calls to be switched between the
public switched telephone network and emerging ATM or IP
networks. It provides a comprehensive solution for the
control of MGs by MCGs. As with earlier generations of media
gateway control, Megaco is based on the principle that all
call processing intelligence resides in the MGC. The MG
provides only the capability to connect media streams under
the control of the MGC, and to detect and transmit
signalling associated with those media streams. Connections
within the MG are created via Megaco signalling commands.
The Megaco protocol is designed to be transport independent.
Therefore, methodologies exist to transport the Megaco
Protocol over TCP/IP or UDP/IP in IP-based networks, but
there are good reasons to use an ATM-based transport such
AAL2 to support remote MGs that operate with VoATM
connections, as in the case of VoDSL.
The Megaco framework is a joint activity of the International
Telecommunication Union and the Internet
Engineering Task Force (IETF).
Let's take a look at how VoDSL will likely change as it
migrates to this next gen architecture.
VoDSL in the Decomposed Architecture
As the PSTN evolves toward a softswitch architecture, we can
expect to see VoDSL Access gateways evolving to support
three key developments:
- Support of signalling using Megaco/H.248, enabling the
call control functions to be supported by a media
gateway controller (and therefore eliminating the need
for the classical Local end office switch);
- Establishment of direct packet voice paths between
- Support for direct connections between IADs and
Because the IAD takes certain types of inputs (e.g.,
POTS, LAN) and transforms them, by adapting the traffic
using ATM adaptation layers, it can be seen as a media
gateway. Signalling -- in the form of MCGP or Megaco/ H.428 --
performs the call control. Now, the VoDSL gateway has also
been transformed into a media gateway.
Traffic coming out of this media gateway can be routed
either on IP or ATM to the packet network (the NGN).
In this architecture, the IADs have the intelligence to
send/receive info from the softswitch, so they could
exchange signalling messages with it, but never actually
route any signals to a TDM network -- voice packets could
just go from one IAD to another. The benefit is this: The
traffic is already in packets, and stays in packets, and
doesn't incur the overhead (and consequent delay) of
The distribution of the next-generation functions can
take many forms. The general tendency is to evolve the
access equipment to become multi-services nodes (DSLAMs,
Digital loop carriers, etc.) supporting multiple types of
media streams (packet and TDM), apply the proper class of
service parameters and to work under the control of a MGC.
One might tempted to call these platforms NGDLCs -- or Next
Generation Digital Loop Carriers.
Of course, the VoDSL gateway will still be able to
connect to the legacy Class 5 to keep the backward
Other important areas demanding attention on the road to the
converged network, including the migration of VoDSL, are
complying with legal requirements for E-911 and making it
possible for law enforcement agencies to "wiretap"
conversations, even if they are all sliced up into packets.
In the U.S., the latter is referred to as CALEA
readiness, after the U.S. Communications Assistance for Law
Enforcement Act (CALEA). Enacted by Congress in 1994, CALEA
establishes a September 2001 deadline requiring that
telecommunications carriers' services and systems conform to
standards that permit law enforcement officials to conduct
legally authorized electronic surveillance. This capability
involves the ability to duplicate, redirect, and fork media
streams so that applications can redirect media streams to
media servers for voice processing applications. (Note:
Media streams redirection and forking also allows easier
development of enhanced services such as calling card and
call redirect or transfer applications.)
Other areas that may not be trivial in the pursuit of
network convergence are the ability to bill actual services
rendered as well as all the Operations, Maintenance,
Administration, and Provisioning -- essentially all the
carriers' back-office operations. These operations represent
at least one-third of carriers' capital expenditures.
Therefore, having technical solutions replicating the
functionality of the PSTN solves only a small portion of the
problem. The operational shift will be equally challenging,
if not more than the convergence of voice and data traffic
over packet networks.
Back to the Future
For many service providers today, VoDSL is an important
technology that can provide an excellent source of revenue.
Today's topology achieves convergence in the local access
part of the network while leveraging existing network
deployments: The ever ubiquitous copper pair and the PSTN
network, as well as DSLAMs, which are seeing continuous new
installations as well as being already deployed in thousands
of central offices. Today's VoDSL implementation also
leverages ATM, a well understood and developed set of
standards that provide, among other key parameters,
end-to-end class of service.
As we move towards the converged networks of the future,
decomposing the functions of the central end-office into
multiple, standards-based functions, VoDSL can evolve nicely
to become a pure media gateway controlled by a media gateway
controller, providing a smooth migration path.
While there are challenges in this migration, the
industry momentum and the spirit of the Internet -- openness
as opposed to proprietary protocols -- will prevail. It is
just a matter of time.
Frederic Dickey is the Product Management Director of Natural
MicroSystems' Network Access group. Natural MicroSystems
is a technology leader supplying the world's major
networking and communication equipment suppliers with the
products and services required to create communications
infrastructure, applications, and services faster and at
lower costs than ever before.
To The June 2001 Table Of Contents ]