HARDWARE

It almost doesn't matter what hardware you use.

The ideal way to select the hardware for any computing task is to first ask yourself what software you need to satisfy the task. The software should drive the hardware. In the real world, this is not always the case. Consequently, you may not have to ask yourself what hardware to use. Rather you may be asking yourself, "What hardware is available?"

If you do have a choice of hardware, then first articulate and enumerate the purpose(s) of your server. Analyse who your audience is and try to guesstimate how many hits your server will get per day. If your server is intended to disseminate lists of services or act as a conduit for your usual print publications, then a microcomputer-based server will do just fine. If you are planning for a departmental Intranet, again, a microcomputer-based server will fit the bill.

Consider exactly what any HTTP server is doing. It is disseminating small (maybe 35 K) files. Thirty-five kilobytes of data is tiny. It does not take a large computer to serve this amount of data over and over again. Furthermore, it has been proven microcomputer-based servers can handle thousands of hits per hour, if not tens of thousands of hits. The use of your microcomputer as a server may also be determined by what information in databases you have that you would like to share. If your data resides in a microcomputer, then you will be using the same microcomputer to serve this data (unless you want to move it to another machine).

In short, microcomputer-based servers offer a number of distinct advantages. First, the use of an operating system you are already familiar with; you will not have to learn something like Unix and all of its administrative overhead. Second, the microcomputer hardware is readily available. Starting out, consider using one of those computers in the back room that may be gathering dust.

On the other hand, the use of a Unix, Windows NT, or even a VMS-based server may be something to consider. This is especially true if you desire to serve the data from your OPAC through a WWW interface. These multiuser, rock-solid operating systems come at a cost. One is administrative support. You will almost necessarily have to have some sort of network "guru" managing your system. For you there may be a learning curve while you get familar with the Unix, Windows NT, or VMS environment. Another cost is distance. "Out of sight, out of mind." In other words, if your HTTP resides on a computer in the back room that only a few people have direct access to, then the computer turns into a mystery. This is not the best way to perceive of computers and makes them seem far away and out of your control.

At the same time, these bigger computers were designed to run client/server applications. TCP is a fundamental part of the Unix operating system, and consequently there is a lot of support for TCP networking built in. Also, there is no denying it, these bigger computers are faster. But remember, the speed of any network connection is only as fast as the slowest link in the chain. If you users are all using 28.8 modems, then it doesn't matter whether or not your server is a blazingly fast minicomputer.

In the end, you will probably find yourself having an mixture of micrcomputer-based and minicomputer-based HTTP servers in your institution. When you are just learning, use your desktop computer as a test bed and grow from there.