Before reading this you may find the following tutorials useful:
Protocol Basics, Formatting for Transmission, Internet Protocol (IP) Life is full of compromises. The telecommunications industry is quite life-like in this respect – ideal solutions are a rarity and so compromises often need to be made. Out of the range of protocols available, it is perhaps the role of Asynchronous Transfer Mode (ATM) to keep everyone happy.

ATM is a data-link layer protocol like Ethernet, aimed at wide area networks (WANs) as well as local area networks (LANs). Whereas Ethernet is really geared towards carrying only Internet Protocol (IP) traffic, ATM is designed to integrate both data and voice needs in one network.

If you can imagine your government outlawing all legal tender apart from one single coin, you may have an idea of how ATM works. Picture life with just one coin, perhaps worth enough to buy, say, a pair of shoes. You just have a supply of this one coin, and you must use that to cover all of your money transactions.

ATM has fixed-length “cells” of 53 bytes in length in contrast to Ethernet’s variable-length “frames.” It is this size of cell that is a compromise between the large frame requirements of data transmission and the relatively small needs of voice calls – just as your one denomination of coin is a compromise between the small amount of money needed to buy a cup of coffee and the relatively large amount required to buy a television.

By catering to both forms of network traffic, ATM can be used to handle a company’s entire networking needs, removing the need for separate data and voice networks. The performance, however, can also be compromised, and the network may not be as efficient as dedicated networks for each service. Try to buy a house with a bagful of coins and you should realize the problem.

ATM’s 53-byte cells each contain a 5-byte header along with 48 bytes of data, and overall transmission speeds from a few tens of Mbit/s up to several Gbit/s are possible. ATM is a connection-oriented protocol, meaning that it must establish a connection (or “circuit”) between two devices before it can begin to send information between them. This is different from the working of Ethernet where frames are sent out with addressing information to find their way around (so-called “connectionless” networking).

The benefit of the fixed-length cells is that the traffic in the network is very predictable, and very fast switching of cells is possible due to the inbuilt clocking of cells travelling through. Due to this predictability, it now becomes possible to guarantee specific levels of service to traffic that may be very time sensitive (such as voice calls and video conferencing). IP networks using Ethernet are what is known as “best effort,” meaning that no guarantees of quality of service (QOS) can be given, and the network will just do its best to get information from A to B as quickly as it can. In this respect, ATM really has the edge.

When ATM sets up a connection between two devices, it uses what are known as virtual channels (VCs) and virtual paths (VPs). A VP is a route between two points, and within it can be many (up to several thousand) VCs that are initiated for each specific information exchange. The advantage of having the paths in place is that all channels within a path can benefit from the same management functions. An additional plus is that adding new channels is simpler, as the foundations of the link are already in place. Once such a VC is in place, cells can be switched relatively quickly through intermediate points and direct to their destination.

The 5-byte header in the ATM cell contains a virtual path identifier (VPI) and a virtual channel identifier (VCI) that serve to identify the exact path and channel each cell belongs to – forming a “virtual circuit.” It should be noted that there are two types of connection that can be made. Permanent virtual circuits (PVCs) are connections that are constantly in place, whereas switched virtual circuits (SVCs) are those that can be dynamically created and destroyed as required by the network.

The final piece of ATM-speak we shall be covering for now is that relating to ATM Adaptation Layers (AALs). AALs convert information into ATM cells, and are used to determine the QOS that specific traffic will receive. The four main types of AAL you may hear about are CBR, VBR, ABR, and UBR. Constant Bit Rate (CBR) gives a fixed and guaranteed amount of network capacity to a circuit, and is therefore useful for time-sensitive information such as voice. Variable Bit Rate (VBR) circuits have the ability to grow and shrink in size with their data needs, but some loss of cells may occur if requirements cannot be met due to higher priority traffic from other sources. Available Bit Rate (ABR) guarantees a minimum level of data transfer with the possibility of higher rates if available. And finally, Unspecified Bit Rate (UBR) circuits just provide whatever spare capacity the network has at any one time.

As a final note, in a real-life application it is likely that IP traffic may be travelling through an ATM network that may then go out onto optical fiber through Synchronous Digital Hierarchy (SDH). Such a scheme is referred to as IP over ATM over SDH.

Key Points

• A data-link layer protocol designed to support voice and data
• Fixed-length "cells" of 53 bytes (including 5-byte header)
• Connection-oriented protocol with predictable network traffic due to fixed-length cells
• Quality of service is possible, in contrast to Ethernet’s "best-effort" service
• Virtual paths (VPs) containing several virtual channels (VCs); identifiers for each (VPI and VCI) in the headers define a virtual circuit
• Permanent virtual circuits (PVCs) and switched virtual circuits (SVCs) are possible
• Adaptation layers provide different levels of service from Constant Bit Rate (CBR) through to Unspecified Bit Rate (UBR)

Further Reading

Sonet (Synchronous Optical NETwork) and SDH (Synchronous Digital Hierarchy)