Network Computingwww.networkcomputing.com

Print this Page E-mail this URL

Several ATM Adaptation Layers are available:
- ATM Adaptation Layer Type 1 (AAL1): designed for constant bit rate, time-dependent traffic such as voice and video.

- ATM Adaptation Layer Type 2 (AAL2): designed for variable bit rate traffic such as compressed voice.

- ATM Adaptation Layer Type 3/4 (AAL3/4): intended for variable bit rate, delay-tolerant data traffic requiring some sequencing and/or error detection support.

- ATM Adaptation Layer Type 5 (AAL5): used for variable bit rate, delay-tolerant connection-oriented data traffic requiring minimal sequencing or error detection support. LAN traffic is typically encapsulated with AAL5, and many compressed voice implementations use AAL5 as well.

Traffic Classes
While the Adaptation Layer conversion is performed at the edges of the ATM network, traffic classes specify the characteristics of the underlying transport. Carriers typically offer different service guarantees based on the desired service class, and in designing your network you'll need to map each traffic stream to the desired class. It simplifies logical topology design to consider each traffic class as its own logical network.

Several traffic classes are in common use:
- Constant Bit Rate (CBR) service supports a constant or guaranteed rate to transport services such as video or voice. It's also used for circuit emulation services (CES), which offer services designed to act like existing TDM services. CBR is reliable, with rigorous timing control and performance, but is less efficient than other traffic classes.

- Variable Bit Rate (VBR) service provides connection-oriented services, using peak and average traffic parameters in order to support variable traffic. It is broken down into two sub-classes:

- Variable Bit Rate-Real Time (VBR-rt) is used when controlled delay is essential, such as real-time compressed voice or video.

- Variable Bit Rate-Non Real Time (VBR-nrt) does not offer delay controls.

- Available Bit Rate (ABR) service supports connectionless protocols like IP, and provides explicit rate control feedback. ABR doesn't specify delay explicitly.

- Unspecified Bit Rate (UBR) service is for best-effort delivery of low-priority traffic.

If you're going to be using a public ATM service, it's important to note that not all carriers offer every traffic class, and the design parameters used by each carrier may be different. You should understand the way the service provider interprets the definition of each class, and the presence or absence of a service class offering may not be a problem. For example, a carrier's published latency and jitter specifications for VBR-nrt service may be perfectly adequate to handle compressed voice and video.

Permanent Virtual Circuits vs. Switched Virtual Circuits
Virtual circuits are the logical transport 'pipes' across the ATM backbone. When using Permanent Virtual Circuits (PVCs), the VC is administratively defined, specifying the complete path across the network. This can be an arduous process in large networks, and doesn't offer the extensive, dynamic rerouting capabilities found with Switched Virtual Circuits (SVCs).

Using SVCs, the circuit is brought up dynamically using ATM signaling. The principal benefit of using SVCs in your network design is to allow direct connection between sites where permanent PVCs would be too expensive to provision. For example, intelligent access devices can setup SVCs based on the digits dialed from a voice handset, eliminating tandem hops in your voice network and saving PBX ports and access circuit congestion, while at the same time reducing call setup time. Likewise, delay-sensitive packet-based traffic such as SNA or Voice over IP can benefit from using SVCs.

Unfortunately, few carriers offer SVC services to customers, although many use them in their backbones to deploy Soft PVCs (SPVCs), launching SVCs based on traffic at the customer PVCs. Many of these same carriers will offer SVCs on an individual case basis, or can tunnel private SVC signaling traffic over PVCs in hybrid public/private networks.

Virtual Paths and Virtual Channels
ATM switches can switch traffic based on a transmission path, a virtual path (VP), or a virtual channel (VC). Virtual Paths are 'bundles' of Virtual Channels see ATMVPsANDVCs to the left, sharing a common traffic class and route through the network. When using a VP service from a carrier, you'll be assigned Virtual Path Identifiers (VPIs) and you'll be free to create Virtual Circuits with any non-reserved VCI value. Using a Virtual Channel service, the carrier will assign VPI/VCI pairs for your equipment.

Traffic Management: Shaping and Policing
In practice, ATM uses much tighter traffic management controls than frame relay, and a mismatch between application requirements and virtual circuit characteristics can break the application. In order to protect one virtual circuit from overrunning the access circuit and WAN backbone, your equipment may shape traffic by buffering cells in order to better match the characteristics of the VC carrying them. It's also been shown that proper traffic shaping is critical to efficient TCP operation.

Likewise, the carrier's network may police cells by tagging and discarding those that don't meet the traffic contract.

Access Interfaces
The ATM User-Network Interface (UNI) specification defines the interface point between ATM endpoints and a private ATM switch, or between a private ATM switch and a carrier ATM network. UNI is the most frequent interface type used to connected a router, switch, or other CPE device and a carrier's ATM switch. However, two alternatives to UNI may be used to access the carrier network: ATM DXI and ATM FUNI. Both protocols are designed to use existing, low-cost frame-based equipment such as routers that may not have affordable ATM interfaces. see FUNIandDXIto the right.

The principal advantage of FUNI over DXI is that frame traffic remains intact over the access loop, resulting in lower overhead. In addition, DXI does not support management of the access interface via the Integrated Local Management Interface (ILMI) protocol, and is more prone to link failures.