What's your network like? How big, how remote, and how varied? How do you keep tabs on it all? Here are a few answers to these questions.
Russell Kay, Technical Editor
No computer is an island, complete unto itself. Not any more. Virtually every computer used in business today is part of a network, and those networks are almost invariably connected to other networks. Most computer users would be seriously hampered in their work and productivity if they couldn't tap into network resources and communicate with other users on their LANs and WANs.
One result of this situation is the increased importance placed on managing an organization's network efficiently. This involves many critical tasks. First, you need to make sure that the various elements of the far-flung web are interconnected properly
and operating the way you expect them to. Given the variety of equipment used to connect network segments and the various protocols that may have to be supported--TCP/IP, Novell's IPX, Apple's AppleTalk, IBM's SNA (Systems Network Architecture), and Digital Equipment's DECnet--this job is far from trivial.
A second important function is administering the normal network business (e.g., rights and privileges) for a user population that is constantly changing and usually growing. Doing this properly, with consideration for security and job-related needs, requires significant resources and a commitment from the organization.
Finally, needs change. In addition, it is the job of network managers to keep track of what's going on in the world of network design to be able to ensure that their network's design (not just its components) is capable of doing what users require. Internet connectivity, for example, is quickly becoming a major new factor in assessing the network's overall load and suitability.
Managers must make significant decisions as to when to limit connectivity, when to expand it, and when the network may require a complete overhaul or redesign.
Picking Up the Pieces
Networks are made up of connected computers and other devices. The connections are made and maintained by still other devices and black boxes--bridges, concentrators, gateways, hubs, multiplexers, and routers. And because the network can have thousands of nodes and range across continents, if a problem occurs you can't just walk down the hall to investigate.
Today's network managers rely on a variety of tools--primarily software--to find out what is and is not happening. The first network management software was proprietary in nature, and for every different type of device--perhaps even for every different vendor supplying those devices--you required a different package with its different interface, training, and so on. That impossible situation is being cleaned up with a new breed of network management platfor
ms that let third-party software feed into a common database and interface.
In ``Monitoring Your Net,'' Paul Korzeniowski describes the development of these tools and shows where the industry is heading. The three major players in the development process include SunSoft, IBM, and Hewlett-Packard. Korzeniowski has assembled a healthy dose of user input as to how well these systems work.
Faster, Faster, Faster!
One issue that network managers hear about time and again is performance. Users are increasingly conscious of, and unwilling to put up with, delays in getting data across a network, whether it's from a file server in the next room or from an internetwork connection halfway around the world. Traditionally, the answer to performance issues has involved faster servers and higher-bandwidth network architectures.
Brett Husselbaugh offers a somewhat different take on the performance question. He believes that bandwidth is not the best measure of network performance and can be mislead
ing. In ``(Mis)using Bandwidth,'' he presents a way of modeling your network mathematically so that you can determine in advance the performance impact of proposed changes to the network.
One key consideration in analyzing network performance, he contends, is that most network accesses are relatively small read requests, and they carry an inherent overhead that is not accurately reflected by bandwidth measures. He draws some surprising conclusions about the relative efficiency of different types of networks. His model is based on years of experience in the field and has been tried and adopted by many network managers and planners.
And Faster Still (When You Really Need It)
Even with the best of planning and appropriate selection of hardware and connecting devices, it's not uncommon for some networks to get so big and so busy that they just run out of steam. When that happens, the answer may be to move to a data transmission scheme that is designed from the start for high-speed, high-volume
data of varying types and different priorities.
The leading candidate for that seems to be ATM (Asynchronous Transfer Mode), which Peter Wayner described in detail in ``On the Road to ATM,'' September BYTE. In this issue, Mark Juliano explains one of the principal reasons why ATM is so well suited to today's networking needs--its ability to allocate bandwidth, control traffic flow, and guarantee the level of service that an application needs.
``ATM Traffic Control'' describes the techniques ATM hardware uses to solve these nontrivial problems and shows why you may want to consider changing to an ATM-based network. In that article, the text box ``ATM with a French Accent'' describes how France Telecom has recently made commercial ATM networks available throughout France and is extending them to the rest of Europe and across the Atlantic.
When you consider how quickly the world has become networked and is becoming ever more so, the true dimensions of the networking management problem begin
to emerge. This is not a one-time obstacle but an ongoing concern that will continue to grow. We need better tools at all levels of network planning, maintenance, and administration. We'll get them, too, but we need to remember that the target is continually moving.
Network Management
