Voice Over ATM: A Hybrid TDM/ATM Design approach
Implementing ATM as a pure transport and switching method in central office design creates an opportunity for immediate technical migration toward a more scaleable environment, without forklifting the existing
local TDM switch fabrics.
By Michel Laurence and Mauricio Peres
For the last several decades, time division multiplexing (TDM) has been the technology of choice for local and interexchange carriers, who used to build telephony systems that transport digital telephony traffic in 64-kbps channels (also called DS0s). Looking at carrier-class central office (CO) designs worldwide, users typically encounter different architectures, such as central or distributed switching, as well as different
CO internal control structures.
The switching system’s most important aspect is its control structure. Older implementations of CO design used a centralized control system, which normally employed one or more dedicated processors. These processors could handle a high capacity of phone calls, and operated synchronously to allow for redundancy. That level of centralization meant that the control CPUs had to present high reliability and a complex operation. Maintenance software had to be capable of
handling a single point of failure.
More recent CO architectures took advantage of the broad availability of very large scale integration (VLSI) technology that has evolved over the past twenty years. These advantages occurred mainly in the area of low-cost single-chip CPUs, which enabled fully distributed control, where the multiple control functions were provided by microprocessors associated with small terminal or subscriber circuit groups. That feature not only simplified the control software (each
CPU performed specific functions associated with the controlled group), but resulted in higher reliability for the CO, since failure in any one CPU only affected a small number of terminals.
For many years, voice-switching architectures, based on multiple-line card shelves with distributed switch elements, have been deployed in the local exchange domain, serving thousands of subscribers with a couple of shelves. In many of the traditional digital CO designs, it is common to interconnect modules or
shelves, such as subscriber termination, operation, and maintenance. Another common practice is to switch modules via a series of buses containing at least three types of signals: user voice traffic, synchronization, and signaling.
Provided that a nonblocking switching system is available within the CO, these three signal categories can be used to implement the basic CO design functions. They may be used to time an interchange or to switch any incoming phone call from the local area to any of the
available outgoing ports in the system that could be sent back to the local area, or to the next hierarchical CO or tandem switch. In TDM switch fabric design, the voice traffic, and in some cases the signaling information, transports in TDM format between the cards in the same module or shelf.
Some variations of TDM switch design for CO applications have been based on functionally integrated modules. Each module in a shelf (chassis, or cell) provides all the line and trunk interfaces, the TDM switching
fabric, as well as the call processing, control, and signaling facilities necessary to support the essential CO switch functionality. In this case, the TDM interface is used between coder/decoders (codecs) or between pulse code modulation (PCM) trunk interfaces, and control or switch units within the card itself, as well as between the multiple cards in the module.
Increasing the CO’s switching capacity
To raise the capacity or scale the switching throughput of most of the existing
TDM-switch-based COs, some architectures interconnect several chassis or cells together with TDM links in a meshed fashion. This provides a nice linear capacity/cost model that is very attractive in a growth market for operating companies. This concept is illustrated in
Figure 1.
When the number of cells in the chassis reaches a certain threshold, this meshed-type interconnection method reaches its maximum operational capacity. Beyond that point, tandem or
hierarchical switching has to be used. (A special type of chassis or cell provides a large TDM switching matrix. This matrix links all the hierarchically lower chassis or cells using TDM interconnections configured in a star topology.) Several hierarchical or tandem switch modules can then be interconnected in a mesh network to build even larger switches (
Figure 2
).
Again, when the number of hierarchical TDM switches reaches its economical threshold, the meshed
network topology reaches its maximum practical capacity. To build even larger switches, another set of hierarchical switching cells, or chassis, would have to be used.
Looking at the existing local and interexchange carrier plants, the use of hundreds or perhaps thousands of TDM interconnects over the wide area network (WAN) results in high-quality, reliable transmission equipment. This guarantees smooth functioning of the entire network, but also increases implementation costs.
Using multiple TDM
interconnects over WANs can raise serious management challenges. The number of interconnecting wires becomes a major impediment, which motivates manufacturers to investigate other possible solutions.
Changing market environment
As previously discussed, switch designers using the distributed modular architecture are assessing the options for connectivity between CO equipment. At the same time, a new, competitive market environment is imposing new requirements that need to be addressed in the
connectivity solution. Competitive local exchange carriers (CLECs) prefer a modular design that has a more linear cost model when compared to more traditional TDM switch designs. CLECs also require modular designs to support a geographically distributed architecture — where the switch modules can be spread over various distant COs while operating as a unique entity. Finally, with the irreversible voice and data convergence trend, TDM switches are migrating toward multimedia switches. (This trend is
exemplified by the recent market introduction by Lucent Technologies of its “AnyMedia” switch family.) The connectivity solution must allow both geographical dispersion and multimedia traffic transport and switching.
Asynchronous transfer mode (ATM) technology is used in transport and switching services that require Quality of Service (QoS), such as voice communications. ATM ad-dresses issues that challenge CO equipment developers who are designing open, large switching systems for real-time,
constant bit rate traffic, such as voice traffic transported over TDM technology. ATM can be used to create networks that offer voice switching systems that are:
- Scaleable
- In compliance with WAN transport protocol standards
- Transparent to end application distance requirements.
Implementing ATM as a pure transport and switching method in CO design creates an opportunity for immediate technical migration toward a more scaleable environment, without forklifting
the existing local TDM switch fabrics. In very basic terms, smooth ATM migration is accomplished by inserting ATM uplink cards in existing TDM-based switches.
Off-the-shelf high-density, hybrid TDM/ATM silicon switch fabric is now providing switch designers with a very appealing and cost-effective alternative. It not only solves the scaleability problem facing their current TDM switch design, but also allows the design of distributed switches, which are capable of migrating towards multimedia traffic and
services.
Hybrid TDM/ATM switching techniques, focused on protecting the carriers’ existing investment in TDM-based voice communications, are becoming popular subjects in several countries worldwide.
Hybrid TDM/ATM switching
ATM uplink cards can be used within existing TDM modules or cells of CO equipment to increase the bandwidth scaleability of the overall switching platform. In
Figure 3
, a hybrid TDM/ATM architecture is used to
solve the scaleability problem discussed previously in the design of very large TDM switches. In this architecture, each cell or chassis is fitted with high-capacity TDM/data/ ATM bridge uplink cards or network interface cards (NICs).
The basic function of these NICs is to directly bridge the outgoing phone calls from a subscriber module or TDM cell into an ATM virtual circuit (VC), predefined by the user at system initialization. Typically, a distributed CO system over ATM, as shown in
Figure 3
, can implement fixed virtual connections between the various source and destination points in the overall network. In this design option, the TDM/ ATM NICs allow the user to intercept hundreds of phone calls at the TDM level, transform them into ATM cells, and interconnect them into an optical link that can run at various speeds (155 Mbps, 622 Mbps, or even 2.5 Gbps). For example, if 155-Mbps optical links are used between multicell TDM switches, up to 2,000 phone calls
could be placed within that pipe. The choice of the optical fiber speed is dependant on the number of voice channels a particular cell can issue at the same time. If 2,000 different people are simultaneously calling the same destination from the same city (from within the same TDM cell), then a complete 155-Mbps trunk will be used at that particular time.
In general, the NIC cards used in this CO design employ a small number of ICs to implement the voice channel to ATM cell conversion, as well as the ATM
signaling needed between the TDM cell switches.
Figure 4
provides a typical block diagram of such an NIC card and its key functional components: the TDM-to-ATM segmentation and re-assembly (SAR) device, which maps TDM traffic over ATM cells; the data SAR, which maps high-bandwidth packet data over ATM cells (used for signaling); and the ATM physical framer, which maps ATM cells over standard transmission links such as Synchronous Optical
Network/Synchronous Digital Hierarchy (SONET/SDH) facilities. Depending on the TDM bus implemented and the connector space available, thousands of phone calls can be delivered to a single TDM/ATM SAR chip.
One way TDM voice data can be mapped over an ATM cell is to use a format called ATM Adaptation Layer 1 (AAL1) structured data format (SDT). This format is specified by the International Telecommunications Union recommendation ITU.363.1, and the ATM Forum Circuit Emulation Services (CES) specification (
Figure 5
).
Figure 5
also illustrates two key features of transporting TDM traffic over ATM. The first is the capability to build trunks of any size over an ATM link facility. This allows the user to size the ATM VC links, in order to match the instantaneous voice bandwidth requirement between two end points, and to ensure the ATM cell assembly delay required to construct this cell. Because each voice sample carried in the TDM
domain takes 125 µs to be generated, packing one channel into the 48-byte payload of a single ATM VC will generate a packetization delay of approximately 6 ms. Packing 48 channels into an ATM VC trunk will generate a packetization delay of only 125 µs, which is equivalent to the delay normally introduced while using standard TDM trunk circuits.
The second key feature shown in
Figure 5
is important in terms of cost and real estate design. It is the
ability to directly extract and insert individual voice channels (DS0) in the ATM payload, without resorting to any TDM framer device. With AAL1 SDT format, each TDM timeslot byte is aligned with the ATM cell byte payload, and the cell SDT pointer indicates the position of the first 64-kbps timeslot in the cell payload. As a result, the data for each TDM channel is readily accessible in the ATM payload, without requiring the use of any kind of TDM framer.
For illustration purposes and cost comparison
analysis between traditional SONET/SDH-based and ATM-based solutions, assume that two TDM switch modules need to be interconnected with a link that’s capable of carrying about 2,000 voice channels (DS0). At each end, the individual voice channels need to be available on the switch module TDM bus.
SONET/SDH option
To provide such transport capability, a SONET or SDH OC-3 link is used between the two modules. If the SONET OC-3 link is operated in a standard TDM mode, each terminating network
card will require at least one SONET OC-3 terminating framer IC, an STS-3/STM-1 mux/demux IC, and at least three each STS-1-to-DS-1 mapper ICs. Then, multiple framer ICs, supporting the functionality of 84 DS-1 framers, plus all the ancillary processing facilities will be needed. This normally involves a multiboard solution, requiring thousands of dollars of hardware components alone.
ATM option
If the SONET OC-3 link is operat-ed in an ATM mode using the AAL1 SDT format shown in
Figure 5
, the terminating card at each end of the SONET link can be implemented using only two key IC components: a SONET STS-3 ATM framer and a TDM/ATM SAR device. Even if a data SAR component is added to allow the link to simultaneously carry TDM and data traffic, the solution still occupies only a few square inches of board space, and the cost of the off-the-shelf hardware components is less than $300.
Each card is capable of transmitting and receiving over
2,000 full-duplex DS0 connections over a single fiber-optic connection (OC-3). Each fiber-optic link can thus replace over 60 E1 (or 80 T1) equivalent TDM links.
To allow for a flexible voice over ATM switch architecture at the TDM/ATM NIC card, the user can implement dynamic AAL1 voice allocation, which facilitates the setup and tear down of phone calls within an AAL1 VC. To demonstrate this feature, let’s assume that several VC connections can be established between the source and destination
paths within the source CO equipment (New York to LA, New York to Boston, for example). When the overall mapping is done, the result is a group of hundreds or thousands of VCs connecting several cities. For the purpose of this explanation, assume that each ATM VC could transport about forty-eight phone calls at any such time.
The dynamic AAL1 voice allocation is required when callers dial different numbers to different destinations simultaneously, affecting the voice contents of a particular VC connecting
two cities. Forty-eight channels are always available within that VC for transporting voice traffic between cities A and B. The control and establishment of the phone calls that are selected to travel within the VC (cities A and B), are done by the call control software in the CO that directly affects the SAR device settings.
Since the phone calls are delivered to the TDM/ATM SAR device via a TDM bus with hundreds of channels, the dynamic capturing of different phone calls has to be undertaken by the
TDM/ATM SAR device. This is done through the use of a channel grooming function that captures multiple 64-kbps channels from different locations in the bus.
In addition, SAR devices with full support of dynamic AAL1 voice allocation should allow the user to dynamically switch on and switch off different 64-kbps channels within the ATM VC. If this feature is not present in the SAR device, an external TDM crosspoint switch is required to intercept every new call on the TDM bus, and switch it into the
appropriate pre-fixed TDM channel of the SAR device destined to travel in that VC.
New network configurations
As shown in
Figure 3
, the ATM links coming from the CO are interconnected in a star network configuration through nonblocking, high-capacity standard ATM core switches (already in service across many countries). In this configuration, each ATM link supports the full range of interconnection services that are required to build this type of
switching service. Large capacity intercell TDM transport, intercell control and signaling data support, and finally intercell TDM clock synchronization and distribution services are examples of this type of service.
Compared to the traditional fixed link interconnection approach using TDM switching, the use of high throughput ATM switches in the middle of the CO switching network provides an added advantage. The ATM switches allow the operating and control system to dynamically allocate intercell
connectivity bandwidth to match the instantaneous phone call traffic requirements.
In addition, the use of an ATM interconnecting and switching core network allows the switch to migrate and evolve seamlessly from a TDM service switch to a multimedia switch. This transition is shown in the two lower switch modules of
Figure 3
. In this case, the ATM network has been allowed to reach the lowest cell level in the architecture’s hierarchy. In one of the switch
modules, this descent has been made using ATM switches located at the edge of the carrier network (or ATM edge switches). One of the cells in the module has transformed into a multimedia cell, capable of accommodating not only TDM-type line cards, but mixed data/voice line cards supporting such services as xDSL. This versatility also allows the switch vendor to build line cards capable of discriminating between voice and data modem traffic, with the ability to off-load Internet traffic from the public
switched telephone network (PSTN) right at the line-card level. This off-loading capability alone can solve the traffic problem that is currently plaguing the telephone operating companies on a worldwide basis.
As another topology option, the lower multicell module could provide an ATM ring configuration with a ring-to-star gateway, to bring the ATM network to the lowest level in the architecture hierarchy, such as the local loop.
Enhanced services
The benefits of the hybrid TDM/ ATM
approach over traditional TDM include improved price/performance and multimedia support. This method provides an attractive alternative for switch designers to build modular TDM hierarchical switches that are not plagued with scaleability issues.
The hybrid TDM/ATM approach can be readily distributed to better serve the emerging CLEC market. It can be interconnected with adjunct voice processing equipment to provide enhanced services, and can be seamlessly evolved into multimedia switches to offer new
services demanded by the marketplace.
Mauricio Peres is director of Mitel Semiconductor’s ATM program. He holds a B.Sc. In electrical engineering and has worked in that field for 15 years. He represents Mitel in the ATM Forum.
Michel Laurence is president and chief technical officer for InnoMediaLogic in Montreal, Canada. He holds a B.Sc. and an M.Sc. in electrical engineering, and has been involved in telecommunications for more than 20 years.
Puzzled by a network processing design issue?
