TDM-To-IP Evolution In The Metro
BY KEVIN WAYNE WILLIAMS AND SCOTT LUKES
[Go right to Offload Where It Counts]
Few will argue the notion that the "network of the future" will be
driven by packet-based Internet Protocol (IP) services. It is common
knowledge that, somewhere in 1999, packet traffic surpassed that of voice in
our nationï¿½s backbone networks. This phenomenon was facilitated primarily
by the huge investments in dense wavelength division multiplexing (DWDM)
technologies made by long-haul carriers in the past five years. Less clear,
however, is the fact that even though packet traffic dominates that of TDM
traffic, plain old TDM services still constitute 90 percent of the
approximately $200B annual North American service market.
Carriers thus face a quandary: How to leverage and protect the TDM-based
networks that currently comprise a majority of their high-margin services,
while migrating the network towards a future of IP? Nowhere is this
challenge more evident than at the edge of the metropolitan access and
transport networks, which act as on-ramps between the user and the network
A service provider's access network is necessarily complex, and resists
simplification. Generally, efforts to simplify the access network only
generate greater complexity. Not only does the equipment chosen for any
rehabilitation effort evolve during the process, acquisitions and mergers
cause even more problems. For instance a service provider based on standard
time-division multiplexing technologies can acquire a competitor that has
built a network based on ATM. Over time, most networks become a mix of
technologies: Copper, fiber, coaxial, and wireless transmission using analog
and digital encoding with time-division, cell-division, and wave-division
multiplexing. Despite the diversification of the access plant technologies,
TDM technologies account for the vast majority of the existing plant.
One trend that shows no sign of slowing is the movement of intelligent
electronics out into the access plant. It is generally cheaper to place
remote electronics close to the customer to collect signals and provide
service than it is to run individual wires to each customer. This outward
movement began with Layer 1 switching and multiplexing elements: Systems
that blindly switch bits without regard to content. The advent of ATM-based
SONET Ring multiplexors, Digital Subscriber Loop Access Multiplexors (DSLAMs),
and ATM Service Access Multiplexors (SAMs) began a move towards Layer 2
switching in the access plant.
As IP becomes more widespread through the backbone network, and carriers
provide higher-layer functionality at the edge, the access network is under
pressure from both sides to become data aware. Despite this, it must remain
compatible with existing infrastructure and services.
Current Access Networks
The particular mix involved in any provider's network will be based on its
strategy at initial market entry. An established Local Exchange Carrier will
tend to have an access network based on time-division multiplexing
technologies. The copper loops connected to each home are multiplexed and
switched through the established switching and multiplexing hierarchy: DS0s
into DS1s, DS1s into DS3s, STS-1s, STS-3s, etc.
While there are many access topologies, two dominate the metro landscape:
Linear, and SONET ring. Linear topologies, which tend to receive less press
than ring-based access networks, are less efficient than ring networks, and
cost more to deploy. They do, however, dominate the installed base, and will
continue to do so for some time to come.
SONET rings, used heavily by the established LECs, are deployed in both
Unidirectional Path Switched Ring (UPSR) and Bi-directional Line Switched
Ring (BLSR) configurations. UPSR configurations are popular in the access
plant, and normally operate at OC-3 (155 Mbps), OC-12 (622 Mbps), and OC-48
(2.488 Gbps) rates.
The important characteristics of UPSR and BLSR rings include the
- They use time-division multiplexing. A given circuit uses a fixed
amount of bandwidth, even when no active traffic is present. Circuits
are available only in fixed sizes, and moving from one to another can be
- They use occupancy doubling for protection. UPSR and two-fiber BLSR
rings both reserve twice the bandwidth actually needed for the signal,
while four-fiber BLSR actually doubles the amount of fiber.
These characteristics have caused a search for more efficient ways to
build rings. Several vendors are beginning to offer packet-based rings,
which offer several advantages over traditional TDM rings, but not without
Packet-based rings attack the problem from two
- Bandwidth is expensive and must be conserved.
- Customer bandwidth needs are variable, and service must be able to
change easily and quickly.
To achieve these goals, the nodes on the ring use packet-based protocols
to send traffic around the ring. Each node on the ring is effectively a
small router, incorporating a packet routing fabric. Packet rings achieve
their stated goals, but the costs are significant:
- Relatively expensive nodes. The cost of each node on the ring is
higher than a standard SONET add-drop multiplexor. This is due to
expense of packet processing and switching hardware.
- Packet latency/reliability. The use of routing protocols requires that
each node receive the entire packet prior to forwarding it, which
increases the latency.
- No interoperability. An ATM ring multiplexor cannot share the ring
with standard UPSR or BLSR ring multiplexors.
- Circuit Switching: All circuit-switched data must be converted to
packet format (using various Voice over IP protocols). This is much more
expensive than traditional TDM technology.
In the short term, the lack of interoperability, and the high expense of
voice traffic handling, makes relying on packet access rings alone
Adding Intelligence to the Network
Another approach in migrating TDM-based networks to IP is to add more
intelligence to SONET ring nodes. This concept that was first pioneered by
Cerent Corporation (now a part of Cisco), and has now been expanded upon by
several metro equipment vendors.
Instead of creating a completely incompatible ring, bandwidth on a
standard TDM ring can be reserved for packet or cell traffic. The traffic
offered to the reserved bandwidth is statistically multiplexed, while the
remaining bandwidth is dedicated to existing circuit services. Such rings
can be made to work with the reserved bandwidth being allocated for ATM
cells, similar to an ATM ring -- or for packet traffic, using PPP or Frame
The advantage of such a hybrid ring is that it is backward compatible
with an existing UPSR or BLSR ring. For a carrier that has a large installed
base, the attractiveness of such a solution is obvious: It allows the ring
to migrate towards data services while maintaining compatibility with the
installed base of TDM services. The advantages of such rings over standard
rings are many, including the following characteristics:
- Compatibility with existing SONET rings.
- Processing can vary according to data type. Packet connections can be
made with packets, and circuit connections can be made with circuits.
- Protection can be variable, based on data type. IP traffic can take
advantage of the inherent protection of the routing protocols, avoiding
the bandwidth doubling inherent in conventional UPSR and BLSR rings.
Circuit traffic can use standard SONET protection if desired.
Effects on Linear Access Structures
As noted earlier, linear access structures predominate. So what does the
future hold for the linear side of the equation? Nearly the same thing as it
does for rings.
An M13 multiplexor has been one of the workhorses of the access network
for years. Like the standard SONET multiplexor, it is a layer 1 device;
every bit that comes in goes out. Frame Relay switches have included direct
M13 interfaces for some time. A Frame Relay switch that allows for multiple
DS1s in and a single M13 out is not much different than an M13 multiplexor
with integrated layer 2 switching capability. All it lacks is the ability to
perform direct circuit connections for some of the DS1s. A similar argument
can be made for ATM multiplexors.
The same pressures that exist in the ring access structures apply in the
linear segment of the market as well. Just as a ring multiplexor can
dedicate sections of the bandwidth for packets and cells, a linear
multiplexor can devote sections of a DS3 or OC-3 bandwidth for the same
purpose. Groups of DS1s or DS2s can be allocated as Inverse Multiplexing
over ATM (IMA), Multi-link PPP, or Multi-link FR to provide variable
bandwidth pipes while still interworking with conventional DS3 systems.
By allocating specific sections of bandwidth for specific traffic,
significant improvements in bandwidth utilization can be provided while
maintaining interoperability with existing DS1 and DS3 cross-connect
Finally, the power of statistical multiplexing and multilayer switching
can convert that linear multiplexor into an intelligent element that can
make more efficient use of bandwidth. When used to feed ring elements,
highly flexible and fault tolerant networks can be built on the same
infrastructure as today's access network.
The outward movement of intelligence into the access plant can transform the
behavior of the network without requiring major upheaval of its physical
structure. The key is to use nodes that provide true multilayer and
multiprotocol capability. Using IP within a TDM-based access plant can
provide for efficient protection by making use of the inherent self-healing
capabilities of standardized routing protocols. Existing equipment can
interoperate with intelligent nodes, allowing new services to be deployed as
needed without disrupting the existing network. The result is a graceful
migration from the dominant TDM services of today to the up-and-coming IP
services of tomorrow.
Kevin Wayne Williams is chief technical officer and Scott Lukes is
marketing director at at MAYAN Networks. MAYAN Networks was founded in 1998,
to simplify the delivery of services, and to reduce cost and complexity at
the edge of the optical network. Please visit their Web site at www.mayannetworks.com.
To The March 2001 Table Of Contents