[Printout | Contents | Search ]

Crossing the client/server threshold

By Robin H. Beck, Tessa L. Bocage, and James M. Choate

The quiet debut of the new PENNcard system early in 1993 marked a milepost in the history of computing at Penn. The system was developed by University Management Information Services for the PENNcard Center to support the Penn ID card process, but it was also the first administra- tive example of the client/server computing model--a model now viewed as the way of the future for Penn's administrative computing environment.

Before PENNcard, Penn's administrative computing environment was a mixture of three computing models--standalone, terminal-to-host, and LAN (local area network) computing. Each of the three models tended to be used in idiosyncratic ways to handle a specific set of needs. Even when all three models were used in the same office, they typically didn't interact much--if at all. Penn's client/server future lies in the judicious melding together of these previous computing models. To understand what this means and how it will affect you, it helps to look at each of these models and examine how the client/server model differs.

Previous computing models

Underlying every administrative system previously developed at Penn is the terminal-to-host model of computing. In this model, the host computer is typically a large, powerful mainframe or mini-computer. The application software and the data it creates and uses are physically stored on the host computer; the application and the data can be accessed by devices attached to the host through a network. These devices, even when they are desktop computers, are merely acting as "dumb terminals." That is, once connected, they require no intelligence or memory of their own because all the real work takes place on the host computer. The processing power, sophistication, and capacity of the host can support a rich database environment and a multitude of concurrent operations. In this scenario, multiple users can simultaneously share a single software application and its associated data but a graphical user interface is simply not possible.

As Penn moves to a client/server environment, the usefulness of dumb terminals and low-powered desktop computers will quickly come to an end.

The LAN computing model is basically a hybrid of the other two. At Penn, LANs are most commonly configured to allow file sharing and print sharing. In a LAN, multiple computers are linked together through a network; application software and data typically reside on a single computer called a "file server." The application software is physically stored on the file server and when a LAN user needs it, the file server sends the user a copy of the software so that the software actually executes on the user's desktop machine. The data needed by the software can be stored either on the user's desktop machine or on the server-- where it is available for others to use. Thus, this model begins to take advantage of the strengths offered by multiple computers working together. However, in general, file "sharing" can only occur serially; that is, a single file cannot be accessed by more than one user at a time.

Rethinking the models

If part, or all, of the presentation were performed on your workstation and the other two tasks were accomplished on a network host, you would have an example of client/server computing. At Penn, the Macintosh TCP/IP PennInfo software exemplifies this particular division of task allocation. There are other ways to divide the three tasks in the client/server model--just move more tasks to the desktop machine. In the PENNcard system, the desktop machine is responsible for two tasks--presentation and application logic--and the host performs only data management tasks.

The important point to remember is that in client/server computing the three main components of a single software application-- presentation, application processing logic, and data management-- execute, either wholly or in part, on different computers. In this split, one or more computers "serve" the needs of a "client" computer. Typically, the desktop computer is the "client" and one (or more) network hosts act as the "server." Thus, in client/server computing, you start to hear terms like "database server" (which is very different from the "file server" concept found in simple LANs), "application server," and "display server," depending on how the application's three tasks are split across platforms.

Regardless of how the tasks are split among client and servers, the desktop workstation is a critical component in this model. At a minimum, the workstation must be powerful enough to support a sophisticated display manager. As Penn moves to a client/server environment, the usefulness of dumb terminals and low-powered desktop computers will quickly come to an end.

A closer look at PENNcard

But it was not just the benefits of GUI that drove the selection of the client/server model for PENNcard. To create individual ID cards, the PENNcard application had to be able to control other devices--a card encoder and printer--attached to the workstation. This meant that, for all practical purposes, the application itself needed to run on the desktop. The data, however, had to be available to other applications running on computers attached to the campus network, PennNet. For instance, the data used by PENNcard is shared by the PennNet Authentication System, by the online e-mail directory update facility, and by the account creation processes used by the central e-mail machines (Pobox and Dolphin). Before PENNcard's implementation, each of these systems would have had to maintain a series of cumbersome procedures to capture the same information. The PENNcard system not only simplified the data collection and data sharing process, but improved data accuracy at the same time.

The technology is improving so rapidly that it's no longer a matter of "should we or shouldn't we" but rather "when and how will we" implement the client/server model.

Reaping the benefits of client/server, however, is not an overnight affair. For the PennCard Center, the GUI interface has provided a tremendous improvement but it needed to be introduced gradually so that staff could become comfortable with a windowing environment. The power of the desktop and the complexity it controls also introduce some constraints that weren't there before. The machine on each user's desktop is no longer simply a "personal" computer that can be configured according to individual preferences and needs. It is now an integral part of a much larger and more complex computing environment that transparently combines aspects of the standalone model with aspects of the other two. To be productive, staff need careful training and comprehensive support for all of their desktop computing questions and problems.

Moving forward

Client/server enables the integration of external information and existing personal productivity tools (such as word processors and spreadsheets) into Penn's mainstream administrative applications and furthers the business objectives. We envision an environment in which University administrators and staff will be able to "point and click" their way through administrative tasks such as purchasing, payroll, budgeting, general accounting, and student records, and to easily access, manipulate, analyze, and share the information they need for day-to-day operations and planning.

Client/server offers enormous opportunities for improving the way we do our jobs and the way we create, use, and share information across the campus, but it can be complex to implement and to manage, particularly in a decentralized, heterogeneous computing environment like Penn's. Fortunately, the enabling technology is improving so rapidly that it's no longer a matter of "should we or shouldn't we" but rather "when and how will we" implement client/server. Clearly, the landscape of administrative computing is changing and the future holds many challenges as well as opportunities. But through the ongoing work of efforts like Project Cornerstone and the experience gained from applications like PENNcard, Penn has already made significant progress in positioning itself to tackle these challenges and to take advantage of the opportunities client/server presents for the future.