Give 'em an A...

For many telephony providers, the original rationale behind implementing asynchronous transfer mode networks was so they could handle voice traffic. Yet only after the emergence of standards-based capabilities have carriers been able to deliver on the ATM promise.

The first breakthrough was ATM adaptation layer 1 (AAL1). Standardized by the International Telecommunication Union and ANSI since 1993, AAL1 is now incorporated in the ATM Forum specifications for circuit emulation services.

In the development of ATM standards, AAL1 found its niche as a way to allow ATM to replace time division multiplexing (TDM) circuits at fixed rates such as T-1 (1.54 Mb/s). Its use was subsequently extended to allow replacement of traditional digital voice circuits, providing a means to convey voice on ATM backbones instead of TDM infrastructures. Although not optimized for voice-over-ATM applications, AAL1 was a de facto standard in the absence of a real specification.

But while AAL1 is quite appropriate in some cases, it is not an optimum solution for voice over ATM. To overcome the excessive bandwidth needed for structured circuit emulation, and to provide the flexibility that allows the network operator to control delay on voice services, the industry required a new adaptation layer. ATM adaptation layer 2 (AAL2) was created by the ITU-T with the specific mandate to provide efficient voice-over-ATM services.

AAL2 is a variable bit-rate (VBR) service, as opposed to AAL1's constant bit-rate (CBR), which suffers from permanently allocated bandwidth that is poorly and inefficiently used. Recognizing that AAL1 was not an optimal solution, some equipment manufacturers have developed proprietary solutions that increase network efficiency. These solutions have some benefits but may also negate interoperability. Instead, much of the ATM WAN community has opted to participate in the development and implementation of AAL2 VBR voice services, which not only provide support for an optimum VBR voice-over-ATM solution but also ensure network and product interoperability.

Before AAL2, users wanting to implement voice over ATM had to live with the limitations of AAL1 or go proprietary. With AAL2 reaching technical completion, ATM network planners now must consider it for VBR voice-over-ATM appl ications.

VBR voice over ATM AAL2 supports the following features in addition to the AAL1 protocol:

* Efficient bandwidth use through VBR ATM.

* ATM bandwidth reduction support for voice compression, silence detection and suppression, and idle voice channel deletion.

* Multiple voice channels with varying bandwidth on a single ATM connection.

AAL2 provides bandwidth-efficient transmission of low-rate, short and variable packets for delay-sensitive applications and makes use of the more statistically multiplexible VBR ATM traffic classes. Therefore, it is not limited to ATM connections using CBR and can enable voice applications using higher-layer requirements such as voice compression, silence detection and suppression, and idle channel removal. The structure of AAL2 allows network administrators to take traffic variations into account in the design of an ATM network optimized to match traffic conditions.

In addition, AAL2 enables multiple user channels on a single ATM virtual circuit and varying traffic conditions for each individual user or channel. The structure of AAL2 also provides for the packing of short length packets into one or more ATM cells and the mechanisms to recover from transmission errors. Compared with AAL1 with a fixed payload, AAL2 offers a variable payload within and across cells. This functionality provides a dramatic improvement in bandwidth efficiency over either structured or unstructured circuit emulation using AAL1.

AAL2's structure, as defined in the ITU-T recommendation, is shown in Figure 1. AAL2 is divided into two sublayers: the common part sublayer and the service-specific convergence sublayer.

The AAL2 common part sublayer, as fully defined by the ITU-T, provides the basic structure for identifying AAL users, assembling and disassembling each variable payload, performing error correction and linking with the service-specific convergence sublayer.

Each AAL2 user can select a given service access point associated with the quality of service (QOS) required to transport that individual higher-layer application. AAL2 uses the service provided by the underlying ATM layer. Multiple AAL connections can be associated with a single ATM layer connection, allowing multiplexing at the AAL layer. The AAL2 user selects the QOS provided by AAL2 through the choice of the service access point used for data transfer.

AAL2's common part sublayer possesses the following characteristics:

* It is defined on an end-to-end basis as a concentration of AAL2 channels.

* Each AAL2 channel is a bidirectional virtual channel, with the same channel identifier value used for both directions.

* AAL2 channels are established over an ATM layer permanent virtual circuit (PVC), soft PVC or switched virtual circuit.

The multiplexing function in the common part sublayer merges several streams of packets, or individual voice circuits, onto a single ATM connection.

In the ITU-T recommendation the AAL2 service-specific convergence sublayer is defined as the link between the AAL2 common part sublayer and the higher-layer applications of the individual AAL2 users.

The ITU-T has developed two service-specific convergence sublayer recommendations. The first provides a service-specific convergence sublayer for AAL2 to handle segmentation and reassembly for data users. The second provides a set of features for voice users of AAL2, which includes compressed voice, silence indication, alarm handling, channel associated signaling, dialed digits and fax demodulation/remodulation. This service-specific convergence sublayer also formats voice as 64 kb/s pulse code modulation-a mandatory feature that ensures interoperability between networks that may be using AAL1.

In addition, the ATM Forum is nearing completion of a specification that incorporates the ITU-T's recommendations.

Trunking efficiency Because of the complexity of dealing with both fixed and statistical compression in a voice channel, a more application-oriented way of viewing the efficiency of an AAL2 connection is to identify how many voice channels may be carried over a fixed bandwidth ATM trunk between ATM network elements.

Figure 2 shows the number of voice channels that can be carried over a T-1 ATM trunk using AAL2. The X axis represents the value of packet fill delay; the Y axis shows the number of voice channels carried. Note that when using 64 kb/s pulse code modulation inside an AAL2 frame, a maximum of 18 channels can be supported with a delay of up to 8 msec. For the 32 kb/s voice encoded channels, a maximum of 35 channels may be supported.

If silence suppression is included, the carrier gains significant channel capacity (Figure 3). Assuming a voice circuit contains 50% silence and that 20% of all channels are idle at any one time, the number of 64 kb/s PCM channels supported by AAL2 more than doubles from 18 to 45 with 8 msec. of packet fill delay by adding silence detection and suppression and idle channel removal.

When 32 kb/s voice compression is added, the T-1 trunk can accommodate up to 87 high-quality, multiplexed voice channels. Finally, considering that carriers can use this bandwidth for other applications such as remote server archiving or software downloading, ATM networks emerge as the only viable underlying technology for efficient wide area multiservice networking.

By far, the most important benefit of AAL2 is the ability to substantially reduce bandwidth requirements for supporting voice-over-ATM networks. It not only drives down network costs, but AAL2's inherent flexibility also enables carriers to add features via the service-specific convergence sublayer structure such as different fixed compression techniques and voice channel switching between ATM circuits. The ability of vendors to add signaling control protocols between ATM elements and PBXs for call setups and teardowns provides even greater cost and bandwidth efficiencies and flexibility.

Carriers and corporate enterprises around the world are realizing the promise of voice over ATM. In fact, the multiservice strengths of ATM WANs are now presenting compelling business cases as well as technological elegance.

Although ATM adaptation layer 2 offers significant benefits over ATM adaptation layer 1 in bandwidth efficiency, it should be used only when all network elements within an ATM network support AAL2. If a network element cannot support AAL2, configuring AAL1 connections for voice to this node enables the transport of voice.

In addition to using AAL1 for interoperability with nodes that do not support AAL2, some applications are adequately supported with AAL1 and, in fact, cannot even use the benefits of AAL2. A typical application of this type is interconnection to an Internet service provider (see figure).

Here the user access lines are traditional voice lines connecting with a voice end office, yet they serve a dual purpose by providing access to a public voice network as well as access to the Internet. As deployed, the ATM network is not part of the voice network but rather provides interconnection from the Class 5 voice switch to the two ISPs. The ATM network traffic is entirely modem data, which is adequately served with AAL1 and cannot take advantage of the statistical gain associated with AAL2 voice traffic.

AAL1 is also well-suited for applications in which the ATM network provides the functional equivalent to a digital cross-connect system in time division multiplexing (TDM) networks. Physical connections are made between the PBX, TDMs, channel banks and other traditional voice and data equipment and the ATM backbone. Through structured circuit emulation, the ATM network provides routing for any number of DS-0s (64 kb/s), allowing for interconnection of the voice and data equipment. By configuring virtual circuits within the ATM network, carriers can interconnect to external customer equipment. This type of configuration demonstrates the flexibility of the ATM network, providing for unlimited mesh connectivity.