Asynchronous Transfer Mode (ATM) is a technology that requires a revolution in testing. Actually, ATM embodies two technical revolutions: a quantitative revolution in the sheer capacity of communications circuits, and a qualitative revolution in the kinds of applications that the technology will enable.
Vendors of ATM products are coming to grips with the opportunities and problems that 45 Mb/s to 2.6 Gb/s digital circuits can offer. Switches, routers, cross-connects and even workstation interface cards are exchanging cells happily at ATM bandwidths and are beginning to interoperate according to ITU and ANSI standards and agreements worked out at the ATM Forum.
Even test equipment, with its unique need to preserve information that transmission hardware can use and discard, has caught up in absolute bandwidth, even though ATM is the first widely used technology for which the bit rate on serial circuits exceeds the clock rate of most processors. Wide-area protocol emulator/analyzers are generating and monitoring ATM signals at rates up to 155 Mb/s, nearly 100 times faster than its predecessors.
Especially for protocol test instruments, the second revolution will present a stiffer challenge. ATM is designed to provide fair access for digital communications of any medium to circuits and switched networks of enormous bandwidth.
Some of these media, such as voice telephony and computer data, have shared circuits only grudgingly up to now. Others, such as broadcast television, have used entirely separate transmission paths. Still others, such as interactive video, telemedicine and multimedia conferencing, are just beginning to take shape.
The broad bandwidth and inherent capability of ATM to carry any medium without compromise will enable new applications that employ all of these media and more in increasingly novel combinations. Given ATM’s capacity and flexibility, the technical cost of these new applications will be complexity, and ATM allows and demands unprecedented levels of complication.
The Celling of ATM Services
ATM achieves its flexibility through a very simple compromise between the circuit switching method used in voice telephone networks and the packet switching technique common to data communications. Circuit switching offers uninterrupted use of a physical path for the duration of a call, and allows the transmission of information at the constant rates needed to ensure that receivers can reproduce voice accurately. Packet switching offers extremely accurate and reliable transmission for data communications whose rate of transmission can vary from zero to several million bits per second over short intervals of time.
ATM compromises between telephony’s demand for constant bit rates and data communications’ need for “bursty” transmission by dividing fast digital bit stream into cells of exactly 53 octets (424 bits). Each cell carries 48 octets of payload and 5 octets of protocol information, exactly enough to associate the cell with a specific communication and to switch it between consecutive links on its path from source to destination.
The small unit of transmission allows fair access to services and media that require a constant bit rate while ensuring enough flexibility to accommodate bursts of variable-bit-rate traffic. The uniform size of the cell and the simplicity of the cell protocol allow for extremely efficient switching.
Protocols and the Cell Tax
The 5-octet protocol header in each cell imposes a relatively large (9.4%) low-level overhead on all ATM transmissions. Lightning-fast switching, combined with the inherently high bandwidth of ATM circuits, renders this cell tax affordable.
However, ATM, like its predecessor and cousin technology, Frame Relay, delegates the entire burden of error control, congestion control, addressing, routing and management to other protocols used by the services that are encapsulated in the ATM bearer service. The ATM specifications themselves describe a level of interaction called the ATM Adaptation Layer that divides these services functionally by the requirements for constant or variable bit rate, connected or connectionless service, and the need for synchronization between source and destination, into several service classes. Each class is governed by specific protocols that define the rules for segmenting messages into cells for transmission and reassembling them upon receipt.
A service using a single medium, such as a digital voice circuit or an emulated digital voice trunk, might use only one service class, and consequently, one virtual channel through the ATM network. A service using two or more media, such as a data-conference call combining voice and computer graphics, might require two or more types of service and several virtual channels through the network, each with its own protocols.
All but the simplest ATM communications involve at least two sets of protocols. One protocol is used to signal the establishment and disconnection of a circuit, and the other to carry the actual voice, data or video payload of the service.
Data on the Fast LAN-E
The Local Area Network Emulation User Interface (LAN-E or L-UNI) defined by the ATM Forum is an excellent early example of a complex ATM application involving several services. LAN Emulation’s mission involves only one type of information, computer data, and seems simple at first: to provide a Layer 2 bridged connection among two or more LANs, allowing users at stations on any of the LANs to communicate with stations on any other of the LANs as if they were on the same logical network.
The task becomes complex because of the need to simulate a connectionless, shared medium (i.e., LAN) over discrete, switched (i.e., ATM) connections. In operation, a LAN emulation service requires a minimum of three ATM services, one to establish and maintain circuits, one to carry end-user data, and a third to handle the broadcast and multicast features that belong by definition to LAN communications.
As complex services proliferate, so will new and complex demands on test equipment. For services like X.25 packet switching, or even narrowband ISDN, technicians needed to test only the equivalent of one service in one digital medium to characterize an installation or to localize and diagnose a fault.
With services as complex as LAN Emulation or more so, technicians using a single, protocol-intelligent instrument might need to watch four or five connections using as many different protocols to track and solve a single problem. Test equipment for ATM must be able to decode, display, simulate and analyze as many service-specific protocols as possible, especially the Broadband ISDN signaling protocol, which ATM networks will use for call establishment; the ATM network management protocols, such as Interim Local Management Interface (ILMI); and internetworking protocols such as IP (encapsulated according to Internet Engineering Task Force Request for Comment, 1483 and RFC 1577) and Frame Relay, along with their successor protocols to be designed for mixed- and multimedia services.
As the ATM industry matures, the breadth and depth of protocol coverage and the intelligence with which an instrument reduces and analyzes protocol information will make infinitely more difference than pure processing speed.
About the Author
Nick Lombardo is the Product Line Manager for the INTERVIEWR 8000 Series of protocol test instruments at Telenex. He has more than 12 years of experience in the data communications industry, and represents Telenex on the Technical and Market Awareness and Education Committees of the ATM Forum. Telenex Corp., 13000 Midlantic Dr., Mount Laurel, NJ 08054, (609) 234-7900.
Copyright 1995 Nelson Publishing Inc.
|
February 1995 |