Rethinking the access network:
internettelephony.com
RANDY SHARPE
Residential access networks, designed primarily for voice services, experienced remarkably little
change during much of the last century. Today, however, new factors are having the most
profound effect on the access network since the introduction of loop electronics in the 1970s.
The Internet, the accompanying demand for high-speed access and the technologies developed
to meet that demand, compel network planners to rethink conventional access network design
rules.
Limitations of the existing network--particularly the data performance constraints with twisted
pair--for addressing the needs of the future are becoming increasingly evident to those involved
in the public network. To address these limitations, network planners must reconsider the
access network, shifting emphasis away from traditional definitions of carrier serving areas
(CSAs) and into the emerging concept of broadband serving areas (BSAs).
Evolution of the access network
Loop plant design has varied over the years. Today's network is the result of a myriad of
engineering practices that directly affect the ability to deliver high-speed data services such as
asymmetrical DSL (ADSL). Approximately 15% of access lines have load coils preventing their
use for ADSL services. Revised resistance design rules have been popular with network
planners for central office (CO)-served access lines since ISDN was introduced in the
mid-1980s. Approximately 90% of unloaded access lines comply with these rules. However, not
all revised resistance design access lines adequately support ADSL services.
The layout of loop plant also has changed over the years. In the 1960s, network planners
followed a design plan known as the serving area concept. Under this concept, access lines
served from a CO were divided into several serving areas, with each area containing a single
interface point back to the CO. After the introduction of carrier systems in the 1970s, network
planners developed a CSA design plan for carrier system-fed access lines.
As network planners began moving to the CSA approach, digital loop carriers (DLC) became the
equipment of choice for service providers.
A 1991 Bellcore survey found that more than 60% of DLC loops met CSA guidelines. However,
CSA compliance does not guarantee a satisfactory level of ADSL performance. Table 1
summarizes the revised resistance design and CSA loop plant design rules.
CSAs often are divided into one to six distribution areas (Figure 1). A DLC system's remote
digital terminal provides telephony services to subscribers within the CSA. Cable pair groups
feed a single feeder distribution interface in each distribution area where cross-connections are
made.
Historically, deployment of DLC systems was predicated primarily on an economic analysis of the
relative costs to reinforce or replace feeder cables vs. the cost to install a DLC system. Several
factors influenced the decision to deploy a DLC system, including the distance from the switch,
the nature of the switch interface (universal or integrated, concentrated or non-concentrated)
and the cost, size, reliability and power consumption of the terminal equipment.
These factors have led many service providers to deploy DLC systems, causing the economic
prove-in distance from the CO to decline. Consequently, the percentage of access lines served
by DLC systems is increasing. DLC systems now represent approximately 40% of all access
lines, and the percentage is increasing.
Turn up the volume
It has been reported that in parts of the telecommunications network, the aggregate data traffic
now exceeds aggregate voice traffic, with data traffic growing at a much greater rate. In
residential access networks, where high-speed data services such as ADSL are deployed in
earnest, data traffic soon will exceed voice traffic.
Contributing to the rapid increase of data traffic in the access networks is the growth in online
households, the increase in access rates and the longer connection times for data calls
compared to voice calls. More than one-third of all U.S. households have Internet access. This
level of penetration was reached more quickly than with other comparable forms of media such
as radio, television and cable TV.
The Yankee Group predicts that by 2003, nearly two-thirds of all U.S. households will access
the Internet. In the past few years, available access data rates have increased from less than 56
kb/s to more than 1 Mb/s. According to the FCC, the average usage per access line for local
voice calls is approximately 40 minutes per day. In contrast, typical Internet access connection
times are approximately 40 minutes per call. Clearly, the rapid increase in subscription rates and
data rates, combined with longer connection times, will cause data traffic volume to surpass
voice traffic volume in access networks, as it has elsewhere in the network.
Increasing competition
The increase in demand for data services coincides with increased competition from competitive
wireline, cable and wireless network providers. For example, cable networks now cover 93% of
U.S. households, and almost 50% of those are cable modem-ready. Today, approximately
500,000 households subscribe to telephone services over their cable network. Cable operators
continue to install fiber deep into the network and invest large sums of money in infrastructure.
Rapidly falling wireless telephony prices have sparked explosive growth in wireless usage.
According to Cahners In-Stat Group, there are 100 million wireless phones in the U.S. With the
introduction of 2.5 generation wireless systems, untethered Internet access is becoming widely
available. The introduction of third generation wireless systems in the next few years promises
data rates of 384 kb/s or more.
The opportunity for new revenue from high-speed data services is a catalyst to competition for
new services and the traditional bread-and-butter voice service. To succeed in an increasingly
competitive market, network operators must offer a better value to their customers.
Performance issues
ADSL and other DSL technologies deliver high-speed access over the local exchange carriers'
installed twisted pair facilities. The information-carrying capacity of twisted pair wire is inversely
related to its length and may be severely diminished because of surrounding noise and crosstalk
levels.
Simulations show that under pristine conditions, downstream performance in excess of 6 Mb/s
can be obtained beyond 12,000 feet from the CO. However, in simulations more representative
of realistic conditions, performance is substantially diminished.
A comparison of downstream performance of a twisted wire pair with only white noise present
to a loop with two bridged taps and composite noise shows that 6 Mb/s can be confidently
provided under either condition to approximately 7000 feet. It also shows that only 25% of loops
are shorter than that distance. At 9000 feet (CSA loops), there is an uncertainty factor of two in
expected performance (3.5 Mb/s vs. 7 Mb/s), depending on the loop conditions. At 15,000 feet
(revised resistance design loops) the expected performance is between 0 to 3 Mb/s. This
demonstrates the difficulty of confidently offering ADSL service on the periphery of revised
resistance design loops.
Meanwhile, islands with varying levels of broadband service capabilities are a consequence of
the legacy loop plant. Households within 7000 feet of a CO or remote terminal can receive
maximum full rate ADSL service, and this includes approximately 25% of U.S. households.
Households between 7000 and 11,000 feet from a CO or remote terminal can receive some level
of broadband service, and this includes another approximately 25% of U.S. households.
However, achievable rates are uncertain.
Households beyond 11,000 feet of a CO or remote terminal cannot confidently be provided with
ADSL service, and this makes up the remaining 50% of U.S. households.
Approach for a new access network
The old criteria for determining when to deploy active electronics into the access network do not
consider the implications of high-speed data services, both in terms of the nature of the
equipment installed and how the equipment is used. Requirements for high-speed data argue for
a more wide-scale use of loop electronics and smaller serving areas.
Revised resistance design and CSA leave customers on the periphery of the serving area with
diminished ADSL performance. Under the BSA concept, reducing the serving area dimensions
enables the delivery of maximum full-rate ADSL service or very high bit-rate DSL services for an
increasing percentage of customers. Reducing the serving area size also means that fewer lines
are served in each serving area and that the number of serving areas increases.
Critical to the viable deployment of a larger number of smaller serving areas is an efficient and
cost-effective means to transport bandwidth between the CO and the remote nodes located in
the BSAs. The conventional multiple network element approach used in large serving areas
consists of an optical add/drop multiplexer providing transport for a DLC for voice services and
an adjunct remote access multiplexer for ADSL services. This is not practical from the
perspectives of cost, size, power, reliability, data efficiency or element management. A
well-designed integrated platform can provide a less expensive solution that is more compact,
uses less power and is easier to manage.
Transport considerations
Traditional transport products used in the access network are based on time division
multiplexing. The digital transmission hierarchy (for example, DS-1 or DS-3) dictates the
connections' capacity. However, in many applications, the dedication of bandwidth in this
granularity is limiting and wasteful. ATM provides a more flexible mechanism of allocating
capacity to a connection.
Sharing a large pool of bandwidth across a network of nodes--as opposed to dedicating
bandwidth to each node--can improve network utilization, particularly for bursty data traffic.
When a pool of bandwidth is shared across multiple nodes, it can be made available where it is
needed and reduced where it is not.
Statistical multiplexing is used to share bandwidth more efficiently across a network of nodes.
Statistical multiplexing exploits the bursty nature of data traffic and the fact that as more
connections are multiplexed together the aggregate statistics become more predictable. In this
manner, network resources can be allocated based more closely on average rates rather than
peak rates. ATM is well-suited to statistical multiplexing.
High-speed ATM transport can be an efficient and flexible way to interconnect a network of
nodes. It also maintains the quality of service (QOS) needed for voice services and other
services with service level agreements. The inclusion of a small ATM switch element in each
node in the access network can provide the cell routing, intelligent queuing and connection
management required to maintain QOS across the network. In addition, ATM provides the service
independence needed for a full service integrated access platform.
Another transport consideration is network topology. A tree-and-branch topology, for example,
is a good match to the layout of many access networks and is well-suited to accommodate
unanticipated bandwidth demand by the growth of a new branch off an existing node.
In short, an adaptable network of ATM-interconnected nodes in a flexible tree and branch
topology can be deployed to fit the application, rather than forcing the application to fit the
network. This is critical in today's competitive environment. A principal application of this type of
network is a residential access network that can provide uncompromised high-speed data
services through a network of nodes.
Figure 2 shows a distributed integrated access platform providing telephony and full rate ADSL
services to homes within a number of BSAs. The nodes are deployed in a tree-and-branch
topology using the platform's integrated ATM transport capability to efficiently distribute
bandwidth across the network of nodes. The nodes are located at a convenient access point to
the installed loops--the feeder distribution interface.
All access ahead
The exponential growth in the demand for data services and the incessant increase in access
data rates are influencing the deployment of access network systems. A large percentage of
customers are not on broadband service-capable islands and cannot get satisfactory ADSL
service. If this situation is mishandled, both low customer satisfaction and opportunities for
competitors will result.
To overcome the limitations of the existing twisted-pair network in delivering high-speed data
services, a new approach to the access network is needed. BSAs, with their reduced serving
area size, provide a flexible means of moving electronics closer to end customers. This, in turn,
enables carriers to provide the high-speed services demanded by today's customers.
Randy Sharpe is Director of Advanced Technologies at Pliant Systems in Research Triangle
Park, N.C. His e-mail address is rbs@pliantsystems.com. |
