Asynchronous transfer mode (ATM) is the next-generation switching technology of choice. There is little doubt that ATM will meet the infrastructure requirements of the networks of the future. ATM is coming out of the test-lab phase and is now being seriously considered by more and more companies for implementation in their production networks. Who is really using it and why? What is ATM capable of today and what does it promise for the future? A variety of users are adopting ATM for more than the obvious reasons of high-bandwidth applications and relief of network congestion. For example, the medical community uses this technology to deliver digital x-ray images. Financial institutions are successfully completing time-sensitive transactions as well as efficiently transferring huge quantities of data. In some cases IS managers are investing budget money in ATM in order to future-proof their existing networks. In other situations, as new physical plants are being constructed, ATM networks are being installed to avoid massive overhauls in the near future. Although the number of ATM implementations is expanding, as with most new technologies ATM has experienced growing pains. After being, in some people's opinion, over-hyped throughout the early '90s many vendors could not deliver the promised technology in a timely manner. A lack of industry standards had some decision makers holding back and many users simply didn't yet need the bandwidth. Frame relay used for data applications and relatively inexpensive video-conferencing through ISDN have proven to be popular stopgap alternative to ATM deployment. Inter-operability issues have been raised, particularly surrounding VLANs and LAN emulation as well as the services provide by public carriers. Limited network management tools are an area of concern for those who want to use the same tools to manage all their network devices. But progress is being made. Vendors have announced support for ATM Forum LANE (LAN Emulation) 1.0 and UNI (User/Network Interface) 3.0/3.1 specifications. Many of the most important outstanding specifications which address inter-operability issues were just approved in April by the voting members of the ATM Forum. The P-NNI (Private-Network Node Interface) 1.0 specification, which defines the way in which ATM switches communicate with each other, will allow networks to scale in size. The new Traffic Management 4.0 specification explicitly describes the ATM service classes including Constant Bit Rate (CBR) used mostly for voice traffic; Variable Bit Rate (VBR) designed for video traffic; and the much-anticipated Available Bit Rate (ABR) with its rate-based flow control algorithm for bursty data traffic. With products now being developed to these specifications and others, the market for ATM equipment is heating up. While ATM is being successfully implemented today, there are still some issues and limitations. Enterprise-wide deployment of ATM to the desktop is not a reality for most companies. It is expensive to install new ATM workgroup switches as well as Network Interface Cards (NICs) in all of a company's desktop workstations, and the business case cannot yet justify that expense. For ATM to be a viable desktop technology an Application Programming Interface (API) will need to be adopted and software applications that use the API to take advantage of ATM's Quality of Service (QoS) guarantees will need to be written. The ability to guarantee QoS is still a promise of the future in the solutions of many vendors. The elimination of latency, and the ability for time-sensitive traffic (such as voice and real-time video) and non-time sensitive traffic (such as data) to share the same network infrastructure, may become a reality when ATM service classes can be used to optimise bandwidth allocation and compensate for the changes driven by bursts of data. For now, the desktop still belongs to traditional LAN technologies such as Ethernet and Token Ring. With cost-effective switching solutions available, this is likely to remain the case for a long while. Although there are various ways to migrate to ATM, perhaps the most common course merges LAN and ATM switching. Within this combination of switches, ATM access devices are used for packet-to-cell conversion, thus avoiding 'forklift' overhauls. This is known as LAN emulation and is covered by the ATM Forum's LANE 1.0 specification. LAN emulation allows a smooth migration to an ATM backbone without the expense of changing desktop technologies. Today, businesses are implementing ATM as a backbone technology. The bandwidth capacity of ATM is unparalleled. ATM uplinks from traditional LAN technologies are available at speeds up to 155 Mbps (megabits per second) and inter-switch links can be configured at 622 Mbps. File Servers can be directly attached to the ATM network, where they can take advantage of the system performance. Network backbones must be able to support the aggregation and flow of enterprise data traffic, and this is an area where ATM can deliver now. After analysing business requirements and determining a need to migrate to ATM, companies should begin looking for a vendor who will support them through the transition period. Support should come not just in the form of standards-based hardware. Training and education opportunities must be sufficient. Network design with a clear migration path and a strong network management offering are key support elements. Strong field and telephone support capabilities are also crucial to a successful ATM implementation. Questions to consider include: Can the vendor satisfy all of my networking needs for both traditional LANs and ATM? Can the vendor provide an integrated network management application for my entire enterprise? What are the limitations and value-added features of the vendor's solution? What features of ATM are available to users at the desktop today (without replacing the current LANs)? When deploying ATM in your existing network, introduce it slowly by creating a limited test environment. This test area will allow you to observe and learn what ATM can and cannot deliver in your network today. Use the data gathered to prepare for a smooth deployment. Approach the backbone implementation in phases to avoid problem situation, and limit the initial use of ATM to the desktop to users of specialised high-bandwidth applications. Eventually, three main benefits of ATM technology will combine to result in lower network complexity. Increased bandwidth for high performance, virtual networking capability and broadband integration are benefits that will simplify the network, substantially lowering management costs. This is a major reason why information networks incorporating data, voice and video traffic will become the rule rather than the exception in the future. In the future we will see more widespread use of ATM for increased voice, video and data applications. As intranets, based on Internet technology, are utilised ATM will become more attractive because it allows these technologies, which are graphical in nature, to work together in an optimum environment. With a network infrastructure that supports data, voice and video, multimedia business applications will evolve which make the promise of ATM a reality.