Asynchronous Transfer Mode (ATM)

Asynchronous Transfer Mode (ATM)

Asynchronous Transfer Mode (ATM) is a high-bandwidth switching technology developed by the ITU Telecommunications Standards Sector (ITU-TSS). An organization called the ATM Forum is responsible for defining ATM implementation characteristics. ATM can be layered on other Physical layer technologies, such as Fiber Distributed Data Interface (FDDI) and SONET. The relationships of these protocols to the OSI model are shown in Figure 7.11.

Several characteristics distinguish ATM from other switching technologies. ATM is based on fixed-length, 53-byte cells, whereas other technologies employ frames that vary in length to accommodate different amounts of data. Because ATM cells are uniform in length, switching mechanisms can operate with a high level of efficiency. This high efficiency results in high data transfer rates. Some ATM systems can operate at an incredible rate of 622 Mbps; a typical working speed for an ATM is around 155 Mbps.

The unit of transmission for ATM is called a cell. All cells are 53 bytes long and consist of a 5-byte header and 48 bytes of data. The 48-byte data size was selected by the standards committee as a compromise to suit both audio and data-transmission needs. Audio information, for instance, must be delivered with little latency (delay) to maintain a smooth flow of sound. Audio engineers therefore preferred a small cell so that cells would be more readily available when needed. For data, however, large cells reduce the overhead required to deliver a byte of information.

Asynchronous delivery is another distinguishing feature of ATM. Asynchronous refers to the characteristic of ATM in which transmission time slots don’t occur periodically but are granted at irregular intervals. ATM uses a technique called label multiplexing, which allocates time slots on demand. Traffic that is time-critical, such as voice or video, can be given priority over data traffic that can be delayed slightly with no ill effect. Channels are identified by cell labels, not by specific time slots. A high-priority transmission need not be held until its next time slot allocation. Instead, it might be required to wait only until the current 53-byte cell has been transmitted.

Other multichannel technologies utilize time-division techniques to allocate bandwidth to channels. A T1 (1.544 Mbps) line, for example, might be time-division multiplexed to provide 24 voice channels. With this technique, each channel is assigned a specific time slot in the transmission schedule. The disadvantage of this technique is that an idle channel doesn’t yield its bandwidth for the creation of other channels.

Devices communicate on ATM networks by establishing a virtual path, which is identified by a virtual path identifier (VPI). Within this virtual path, virtual circuits can be established, which are in turn associated with virtual circuit identifiers (VCIs). The VPI and VCI together make up a three-byte field included in the cell header.

ATM is relatively new technology, and only a few suppliers provide the equipment necessary to support it. (ATM networks must use ATM-compatible switches, routers, and other connectivity devices.)

Other networks, such as a routed Ethernet, require a six-byte physical address as well as a network address to uniquely identify each device on an internetwork. An ATM can switch cells with three-byte identifiers because VPIs and VCIs apply only to a given device-to-device link. Each ATM switch can assign different VPIs and VCIs for each link, and up to 16 million circuits can be configured for any given device-to-device link.

Although ATM was developed primarily as a WAN technology, it has many characteristics of value for high-performance LANs. An interesting advantage of ATM is that ATM makes it possible to use the same technology for both LANs and WANs. Some disadvantages, however, include the cost, the limited availability of the equipment, and the present lack of expertise regarding ATM due to its recent arrival.

Two other evolving technologies show promise:

Synchronous Optical Network (SONET). Bell Communications Research developed SONET, which has been accepted as an ANSI standard. As the “optical” in the name implies, SONET is a standard for communication over fiber-optic networks. Data rates for SONET are organized in a hierarchy based on the Optical Carrier (OC) speed and the corresponding Synchronous Transport Signals (STS) employed. The basic OC and STS data rate is 51.84 Mbps, but higher data rates are provided in multiples of the basic rate. Thus OC-48 is 48 X 51.84 Mbps or 2488.32 Mbps.
Switched Multimegabit Digital Service (SMDS). Developed by Bell Communications Research in 1991, SMDS technology is related to ATM in that it transports data in 53-byte cells. SMDS (see Figure 7.12) is a connectionless Data Link layer service that supports cell switching at data rates of 1.544 to 45 Mbps. IEEE 802.6 (DQDB metropolitan area network) is the primary Physical layer standard employed with SMDS, although other Physical layer standards are supported.