Note: Many topics at this site are reduced versions of the text in "The Encyclopedia of Networking and Telecommunications." Search results will not be as extensive as a search of the book's CD-ROM.

ATM is a high-speed network technology that is designed for LANs, WANs, carrier and service provider networks, and Internet core networks. It is a connection-oriented switching technology, as opposed to a connectionless technology such as IP. ATM creates a virtual circuit (dedicated path) between source and destination across its switching fabric. These circuits can guarantee bandwidth and quality-of-service.

ATM's fixed cell size provides performance and predictable traffic flows. Picture a busy intersection. A semi tractor-trailer is attempting to negotiate a tight turn. All the rest of the traffic in the intersection is held up while this happens. Now picture the same intersection where all the vehicles are sports cars. In the latter case, traffic flows smoothly, and even predictably, because there are no traffic jams.

ATM cells negotiate ATM switches with the same efficiency, providing several benefits:

• Cell switching is efficient and fast for the reasons just described.


• Traffic flow is predictable due to the fixed cell size.


• Delivery of time-sensitive traffic (live voice and video) can be guaranteed.


• ATM includes QoS (quality of service) features that can be used to guarantee bandwidth for certain types of traffic.

There has been great debate over whether ATM is better than IP and vice versa. Many people find this debate odd, since the technologies are quite different and not even in the same protocol layer. The battle is really about whether networks should be connection oriented (ATM) or best effort (IP). ATM's fixed cell size and virtual circuit capability makes it the best choice for real time multimedia. Carriers and service providers use ATM in their core networks because it lets them provide service guarantees to their customers. However, IP's simple packet forwarding model has proved its usefulness in the Internet, where traffic is bursty and unpredictable. This model allows millions of people to share the bandwidth of the Internet without setting up virtual circuits in advance. However, the IP model starts to break down under traffic loads and congestion. In addition, the unpredictable delays of IP networks are a problem for real-time traffic.

ATM was originally defined by the telephone companies and has been heavily promoted by them as an end-to-end networking technology, as well as a voice technology. In this respect, ATM is both a LAN and WAN technology that can potentially allow customers to replace their separate voice and data networks with a single network to handle both voice and data, as well as other multimedia content such as video.

In the early 1990s, ATM was widely considered the next-generation networking technology that would extend all the way to the desktop. But broadcast LANs were already entrenched in most organizations and Internet technologies exploded on the scene. And while ATM was hyped for its speed, Gigabit Ethernet (1,000 Mbits/sec) and now 10 Gigabit Ethernet offer cheaper and more easily managed services.

Still, ATM is a viable technology for backbones, even in Gigabit Ethernet environments. ATM is easily scalable and integrates with most existing technologies.

More recently, new technologies such as DWDM (Dense Wave-Division Multiplexing) and optical networking may undo ATM and even SONET. DWDM puts hundreds and potentially thousands of lambda circuits on a single fiber. That means core networks will support very high capacity switched optical circuits, reducing the need for packet switched core networks. Imagine having an entire beam (wavelength) of light allocated for your personal use, switched into place when you need it and taken down when you have finished. That is what the new optical networks could provide. See "Optical Networks" for more information.With those trends in mind, the following sections provide an overview of ATM technology.