Printer Friendly

How to pick the right computer network.

Any computerized office that shares data needs a network. Without it, staffers must carry data-bearing floppy disks from computer to computer--which is not a very efficient operation. But how does an accounting office decide which network configuration to install? What works well in one location may be ineffective in another.

This article describes a process for making that selection for small and medium-sized organizations. It examines and compares the three most popular network types: local area networks (LANs), peer-to-peer networks and shared-processor networks. Networks for very large organizations, such as wide area networks (WANs), which are vastly more complicated, are not covered here.


The networking choice that is best for an office depends on a host of variables, which include

* Resource constraints. How much money, equipment and human resources can an organization reasonably invest in the networking effort?

* Number of user8. Management must consider not only the current number of computer users who will be on the network but also how many users will be added in the years to come.

* Peak usage. A network must be able to support peak demands.

* Processing complexity. Demand for computer processing time varies depending on the nature of the task. For example, a pool of 10 secretaries who primarily perform word processing puts far less demand on the network than two accountants working on spreadsheet programs.

* User locations. The physical distance the network must span influences design choices.

* Security. The network must be able to prevent unauthorized access.


The most popular network configuration is the LAN, in which one or more personal computers (PCs) are the main repositories of all application programs and data files shared by the network users. The central computer is called a file server. While all traffic in the network is managed by the server, all data processing is done locally on each PC. In addition to sharing programs and data stored on the server, users can access each other's data, too. Computers on the network can be connected to a file server in a variety of configurations (see exhibit 1, page 43).

A file server must be sufficiently powerful to handle the work load. A 386 computer or higher is recommended because the server needs at least as much random-access memory and speed as a 386 can provide to manage the traffic from the networked PCs. In today's business environments, a 486 computer and the next-generation 586 may be even better choices.

LANs are complex and require an expert for setup and support. In addition, a network administrator must be able to maintain all the shared software and assign access rights and passwords to users.

Typically, a user terminal (called a node" in LAN jargon) is a PC. However, it can be much less powerful and less expensive (a 286 or even an 8088 computer) than the file server. To communicate over the network, each PC must have a circuit called a network interface card (NIC) installed in it. In addition, special cables must be strung between the file server and each node.

These are the major features of LANs:

* They can handle a relatively large number of users. The most popular network program, NetWare 386, can link up to 250 computers.

* They can accommodate multiple disk operating systems, such as Macintosh, UNIX and MS-DOS.

* Their security features are sound.

* LANs can communicate with other networks, whether they are similar or not.

* Printers can be shared and jobs can be well managed--print queues can be arranged by priority and jobs can be sent to the appropriate printers. These are some of the disadvantages:

* LANs need one or more dedicated file servers, whose sole purpose (with some exceptions) is to run the network.

* Each user node must be a PC, not just a "dumb" workstation--a terminal that lacks computing power.

* LANs are relatively difficult to install and maintain.

* A file server failure can disrupt the entire network, and failure-protection systems are expensive.

* Network operating software is expensive. Some can cost as much as $13,000, and initial hardware (cabling, NICs and connectors) is costly, too.


Peer-to-peer networks are easier to install and maintain than LANs. Each node in a peer-to-peer network must be a PC, an "intelligent" terminal capable of doing its own computing, not just a workstation that is a slave to a central computer. Each PC is connected to its neighbor to form either a single chain or a ring (see exhibit 1, page 43). Each PC can operate on its own or access any other PC in the network. Thus each user node can act as either a workstation or a combination server-workstation. In addition, users don't have to learn complicated commands to access other computers on the network--which is not the case with LANs--making peer-to-peer networks especially popular.

For security, each computer's hard disk can be partitioned in such a way that a portion of its data storage is available to other users and the remainder is dedicated to that terminal only. Thus peer-to-peer networking permits users to share software and data while retaining the privacy of stand-alone applications. Also, backup security can be enhanced by strategically duplicating mission-critical data on several machines so that a single computer failure will not result in a total loss. In addition, peripherals, for example, printers and plotters, can be shared. Peer-to-peer networks provide essentially the same functional benefits as LANs.

Like LANs, peer-to-peer networks require additional hardware--network adapters, connectors and cables--but the hardware is simpler, easier to install and less expensive (it usually costs between $200 and $400 a node).

Novell's Netware Lite and Artisoft's LANtastic are two of the most popular peer-to-peer networks.

Peer-to-peer networks have two major disadvantages:

* No more than about 20 users can be connected.

* If too many users are on the network simultaneously, response time degrades. This response problem can be mitigated to some extent by distributing the shared software and data properly throughout the network.


The third option is the shared-processor network. In this configuration, either PCs or dumb workstations can be linked directly to a central processor (called a "host"). This setup is different from a LAN in that the terminals depend entirely on the host: All of the processing takes place in the host. In LANs, most of the processing takes place in the local machines. Only the storage and retrieval are done by the file server. Thus, in shared-processor networks the local terminals' power is not critical to the overall performance of the network, and the more powerful the host, the better the network performs.

Shared-processor networks require special-purpose software, for which special training is needed. Also, many existing PC application programs designed for single users do not run on shared-processor networks. However, compared with the other two network designs, this type of network produces the best performance per dollar invested.

Popular shared-processor network programs are Digital Research's Concurrent DOS, Software Link's PC-MOS Multiuser, Alloy's 386/MultiWare, Intelligent Graphics's VM/386 and Santa Cruz Operation's SCO UNIX System V/386.

These are the major benefits:

* Shared-processor networks are much easier to install and maintain than either LANs or peer-to-peer configurations.

* They match LANs' and peer-to-peer networks' resource-sharing, security and printer management capabilities.

* The cost of purchasing and installing a shared-processor network can be half what an equivalent LAN would cost and about two-thirds the cost of an equivalent peer-to-peer network.

These are some of the drawbacks:

* A failure in the server halts the whole network.

* The number of terminals that can be attached to the central processor is limited and performance declines as the number of users increases.

* Expansion in the future is difficult and sometimes impossible because of memory limitations.


For help in selecting which network configuration is best for a particular use, exhibit 2, page 45, provides a decision tree that guides a reader through the information maze. It focuses on cost, the number of computers and the complexity of the work the network will handle.

Finding the right network for a particular use is not easy, but the effort is worth the time invested. A well-designed network, tailored to the needs of an office, can be a major productivity booster. But if an office selects one that does not fit its requirements, that same network can become a drain--of money, time and users' patience.


* THE REAL POWER of the computer can be realized only when it's connected to a network. However, selecting the right network for an office environment is not easy. What works well in one office may be ineffective in another. * THREE NETWORK TYPES, all for small and medium-sized offices, are examined: local area networks (LANs), peer-to-peer networks and shared-processor networks. * LANS ARE THE MOST popular configuration. They can handle a large number of users, accommodate multiple disk operating systems and provide good security and printer sharing. But they require powerful dedicated file servers, and each node must be a PC. They also are complicated to install and maintain. A faulty file server can disrupt the entire network. * PEER-TO-PEER NETWORKS are cheaper to install and maintain than LANs. Their commands are not as complex as LANs', they don't need powerful dedicated file servers and they can share printers. However, each terminal in a peer-to-peer network must be a PC and no more than about 20 users can be connected. Also, such networks' response times are slowed by heavy use. * SHARED-PROCESSOR networks require special-purpose software, and many existing PC application programs designed for single users do not run on them. But they are easier to install and maintain then LANs and peer-to-peers, and they cost about half as much as equivalent LANs and about two-thirds as much as equivalent peer-to-peer networks. However, a server failure halts the whole network, and the number of terminals that can be attached to the host is limited. Also, performance declines as the number of users increases, and future expansion is difficult.
COPYRIGHT 1993 American Institute of CPA's
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Planning a Computer Network
Author:Raja, M.K.
Publication:Journal of Accountancy
Article Type:Cover Story
Date:Jun 1, 1993
Previous Article:The proposal postmortem checklist.
Next Article:How one firm became computerized - in many small, well-planned steps.

Related Articles
Communicating with pictures: key to 21st century publications.
Tech stocks get hot.
Protecting the patient: how independent review could force HMOs to behave.
Blown Coverage.
It's our birthday again: we're wiser, thinner and poorer as we turn three, but dammit, we're still here!
No Starch Press.
Greenwood Press.

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters