Connecting Networks

When you want to link one network with others, you're going to need some heavy-duty connectivity. ATM, digital switching, and ISDN are among the candidates you should consider.

Bob Ryan, Senior Editor

Two topics dominate the networking arena today--ATM (Asynchronous Transfer Mode) and switching. You could argue, of course, that because ATM is a switching technology, it is subsumed under the general switching umbrella. Practically speaking, however, ATM is such a powerhouse that it deserves a category unto itself.

ATM is nothing more than a compromise designed to handle two conflicting types of applications and data. The first is regular old data communications--you know, the type of communications traffic generated when you hitch together geographically separate LANs. Such traffic is charact erized by its bursty nature and consequent varying bandwidth requirements. Experience has shown that such traffic is best handled by shoveling the data into packets of varying--and normally large--sizes and transporting them via some sort of switching matrix.

The second type of traffic ATM must contend with is the high-volume data required for multimedia applications such as voice and videoconferencing. This traffic has traditionally been handled by dedicated connections: You dial a phone number and someone answers, establishing a connection. Connections are much more predictable in their bandwidth requirements than are packetized data communications, and they also include a real-time character that is normally absent from data communications. Simply put, if you can't move that video data fast enough to reconstruct a sustainable signal on the receiving end, you're hosed. Dedicated circuits are the best way to handle such real-time data, but they are expensive to operate.

Enter ATM. As with tradi tional data communications, it is packet-based, allowing it to handle bursty traffic. Unlike most other data packets, the ATM packet is a fixed-length cell containing 5 header bytes and 48 bytes of data. (In comparison, X.25 packets can consist of several thousands of bytes.) The fixed and predictable packet length makes it possible to switch ATM cells in hardware--there's no need for each switch to parse the contents of every packet. The fast hardware switching allows ATM networks to maintain the high data rates that are required to support connection-oriented data and applications in a switching environment.

ATM isn't the best of both worlds--standard data communications works best with large packets, while real-world data types prefer dedicated connections--but it is the best available compromise. The hype is not just hype; ATM is a tsunami that will transform the face of networking. To find out when and where the wave will break, read Peter Wayner's article ``On the Road to ATM.''

Despite it s wonders, ATM presents one intriguing difficulty: its name. What are people going to call all those user-unfriendly machines that dispense cash at all hours of the day and night?

Mammals into Dinosaurs

Network+Interop is the most important networking and data communications trade show on the planet. While attending the June get-together in Las Vegas, I heard a number of people refer to the need to ``upgrade legacy LANs'' or ``integrate legacy LANs into future networking architectures.'' Excuse me, but isn't the term legacy an adjective you use to describe lumbering mainframes and archaic SNA (Systems Network Architecture) implementations? When did Ethernet become a dinosaur?

Such questions brought benign (condescending?) smiles from the networking cognoscenti, who patiently explained the shortcomings of today's LAN technology in the face of tomorrow's multimedia applications. No need to rip out such legacy systems, I learned with relief; there's nothing wrong that a liberal dose of switch ing can't see you through. John Bryan explains it all in ``LANs Make the Switch.''

ISDN, Anyone?

The more organizations depend on LANs, the bigger becomes the issue of how you provide remote access to LAN services. Jeffrey Fritz--who in real life provides networking services to thousands at West Virginia University--discusses in the article ``Digital Remote Access'' how digital technology is being brought to bear on the problems of remote access. Remember all those ``Year of the LAN'' proclamations during the last decade? Soon, someone is going to have to declare a ``Year of ISDN.'' It looks like it's finally going to happen.

Digital services are becoming more accessible and ubiquitous, but analog technology (i.e., modems) also has much to offer in the remote-access realm. For example, at Networld+Interop, Dayna Communications announced a new remote-access server that uses PCMCIA modems to connect the outside world with AppleTalk--and soon IPX and TCP/IP--networks. PCMCIA technology eases physical configuration problems and provides hot-swapping to boot.

Setup times make analog solutions inappropriate for LAN-to-LAN connections, but the modem still has its place in single-user access to LANs. Given the advance of digital technology, however, I wonder how long it will be before I start hearing solutions to the problem of ``legacy modems''?

The Collapsing Backbone

The net effect of the new and evolving networking technologies described here is difficult to predict in the short run. Certainly, there will be confusion as router vendors, such as Cisco Systems, invade the hub arena, and hub vendors, such as SynOptics Communications, get into routing. ATM simply adds another flavor of spice to these mixtures.

The long-term picture is clearer. The internetworking functions you need to perform will increasingly take place in a single box. Your multiprotocol network will be multiprotocol in one place only, the switching hub, thus making your network backbone more accessible an d manageable than ever.