This is the second of three articles on micro-mainframe links that will appear in THE SCIENTIST over the next few issues. The first appeared on October 20, 1986.
Author: RONALD F. KOPECK
Date: December 15, 1986
You must give some thought to planning how you want to link personal computers. This does not have to be a lengthy process. In fact, the quicker you assimilate what you need to do, the better. But before you can begin to formulate ideas about how you will link, you need to have a clear understanding of the data needs and activities of your users and the applications and projected growth of your system.
You can develop an initial network from the bottom up or from the top down. In the first approach, you identify specific user needs that you must address immediately, with today's linking technology. In the second, you consider the total system, including the need to expand and modify as your needs change. When you merge the two approaches, you can produce an excellent link plan.
Consider both ends of the program. What do you need to do right now to satisfy users' demands for access? What would you like to do and have operational in the future? A plan that takes six months to implement when users need to get to the host now will not work because you never have the six months in which to get it running. Just when you think you have the perfect link design, you discover that users have changed their information needs and your plan is obsolete.
Develop a micro-mainframe link network by having a master plan and modifying it as you proceed. In this way you stay in touch with the reality of user needs, and you avoid ending up with a mesh of products that do not allow you to develop an integrated system that supports the services and software applications you know are to be resident on your host and available to your users.
Look at new methods for integrating workstations and new products that offer you an opportunity to develop a system rather than just to fix today's problems. All too often, the product a user selects to solve a specific problem does not apply in another, similar set of circumstances. When products are chosen without thought to the system as a whole, micro-mainframe links proliferate without any set pattern.
Look at products based on how they can be integrated with other products. As you evaluate your problems, think of the system as a whole. There are several approaches to linking, each of them with variations. The method you select is entirely dependent on your needs; what is suitable in one instance might not be suitable in another.
In this approach, you supply some individuals in your lab or office with personal computers and permit host access to selected areas or users. Watch carefully what applications are used on these seeds. Know the types of data transferred, the transfer frequency, and the volume to and from the host and note patterns. As patterns evolve, seed other personal computers to see if the patterns are consistent across all the users in the same area.
The advantage of this approach is that it gives you time to observe carefully and without great risk how personal computers can be used constructively. It lets you decide where the systems should be placed and gives you an opportunity to control the dispersement and set the guidelines.
The disadvantage is that you do not know how the system will evolve. The seed generates curiosity and creativity in users, who exploit what the host can offer in ways that can't be predicted prior to host connection.
Here you define existing environment parameters and plan ahead for the complete integration of all workstations throughout the lab or office. Variations concentrate on specific functions or stages.
If you have several hundred or more personal computers in your operation, this can be a monumental task unless the computers are divided into smaller organizational or functional entities. If you have 25 to 250 personal computers, the task is manageable.
The formal design establishes the goals and requirements of the data center for supporting personal computer integration. You understand host resources and their use in the future, what you want users to be doing in the long term, and what applications and tools you expect to provide.
The advantage of this approach is that you identify and define important items. A strategy exists for linking from an overall point of view.
The disadvantage is the amount of time required to develop a formal plan. Personal computers are generally introduced faster than a plan can be assembled, and users sometimes do what they choose while waiting. Often interim solutions become established before a formal design is implemented.
This approach is based on the idea that the solutions that are used to implement host connections are eventually incorporated into an overall network structure or discarded for a different solution altogether.
If you cannot spend a significant amount of time defining requirements and devising overall plans because user demands and requirements dictate immediate solutions, this is the technique you should use. The intent is to get from one point to another. Approaches can be mixed to form stages.
The advantages and disadvantages of this approach vary depending on whether the interim solution is absorbed into an overall plan or thrown away and replaced. If absorbed, the interim provides a usable stepping stone to an end. Calculated investments are made as the solution unfolds. If replaced, the obvious disadvantage is cost and retraining to a new system.
Don't define or implement a complex solution to your problem immediately and don't finesse the issue. Confront the basic problems of integrating across multiple host configurations with as simple and straightforward a solution as possible. You can expect that whatever you design will remain in place for at least 18 months. You should minimize the use of multiple-vendor local area networks or link networks within a configuration. Plan for the possibility of more than one network type but minimize the number of them. This is crucial to ensuring the reliability and operation of the linking environment.
If you buy a local area network, file server or link, the time you spend in identifying potential exposures and operation nuances is a small price to pay given the size of your investment in dollars, time
Ronald F Kopeck is the author of Micro to Mainframe Links, published this year by Osborne/McGraw-Hill, and president of EdgeTech Associates, Suite 602, 582 Market St., San Francisco, CA 94104.