Draft 4/13/97 [Originally written as a future technology overview/direction]
The workstation is the desktop appliance where the knowledge worker interacts with various kinds of information retrieval and processing resources (see Mirror Worlds, David Gelerntner). Our current technology to do this is based on inter-networked workstations and services. The resources to complete a given task may be located on the workstation itself, on one or more servers, or all of these.
To qualify as an "appliance", the workstation must just work. There can be no user design, no user installation, and no user maintenance. This implies full, professional support. Centuries of experience with various technologies shows us that small to moderate centralization is the most effective means.
So then, how do we proceed? The style of the workstation and the style of the central support must be blended to achieve the end user's purpose.
Note that a particular workstation may operate in a different mode for each application. Also a particular workstation may not have enough resources or the specific kinds of resources to act in some modes or for some applications.
The self-reliant workstation with local storage and processing is essentially the "stand-alone" mode. The state and future prospects of the workstation reside in the local storage (and back-ups!).
The self-reliant workstation with mirrored storage has a complete image of itself both locally and on a central server. The two images are automatically re-synchronized whenever the network is operational. The server image need not be a literal copy of the workstation storage, for example programs may be stored separately from writable user areas. "Mirroring" as used here includes such terms as replication, cloning, file synchronization, etc.
The network workstation with cache storage does not maintain local storage of data or programs; its local storage is used only for swapping and caching of server files. Read and write-back cache files are permanently stored only on the server. This is one vision of the "Network Computer" (NC).
The storage-less network workstation relies entirely on the server to provide data and programs. It may cache data and programs in RAM and rely more heavily on programs in ROM, but it is primarily a processing and display engine. This is another vision of the "Network Computer".
The network terminal has neither local storage nor local processing. It is primarily a display device. Today this category includes X-windows terminals, workstations running X-windows servers, and certain proprietary display clients such as for Cytrix WinFrame.
Individual setup and maintenance can be centrally managed using a number of methods and tools. Workstations can be taken to a central location for configuration, installation, and release testing. Roving staff can do routine maintenance, check workstation status, assure that backups have been done, etc. A drop-off depot may be used for later major upgrades, repairs, and the like. All of this requires overall planning, standard configurations, record keeping, etc. The individual setup approach is most often used with laptops or home dial-up systems, but can also be used for fully self-reliant, desktop workstations.
Remote setup and maintenance involves third-party tools, remote control software, Microsoft SMS, etc. The initial installation and more invasive types of work must still be done as for individual setup. Remote setup also requires overall planning, standard configurations, record keeping, etc.
Centralized setup and maintenance of self-reliant workstations requires the mirrored-storage workstation mode. The workstation is managed by working with the server files that represent the workstation. The primary advantages of the mirror approach over the network workstation are performance and autonomy (the mirrored-storage workstation can continue when the network or server are slow or down). The advantage over self-reliant workstations with local storage is that file changes are continually backed-up to the server, and server backup and archive procedures assure long-term survivability and departmental ownership. The initial installation must still be done as for individual setup, but full initial synchronization can then be done from the server.
Setup and maintenance of network workstations and network terminals is managed by the classic methods of the central mainframe. Some additional workstation components and server/mainframe planning and setup are necessary to manage the remote bootup of the workstation.
My current, working preference is for the self-reliant workstation with mirrored storage. This combination seems to maximize the capabilities of all of the technologies involved, including central management.
Powerful, inexpensive workstations are available that are fully capable of self-reliant operation. Use of mirroring rather than full, direct file service makes best use of the shared network resource - especially when some link to the server location is slow or congested. Even the smaller, slower workstations benefit from the mirrored approach.
It does not seem that storage-less or storage-caching network workstations, or methods of server use that operate a self-reliant workstation as if it were just a network workstation, fulfill the performance needs of campus faculty or staff that use workstations. This does not mean that network workstations are not admirably suited to many applications, particularly kiosk and casual use.
While the self-reliant workstation with mirrored storage is the preferred base technology for setup and management, each of the less self-reliant modes may be profitably used when appropriate to the application. Clearly shared data is best served rather than mirrored, particularly if it is being updated. But shared software, system, and workstation configuration is better mirrored or replicated.
Appropriate use of technologies is the highest value, not blind use of maximum or minimum technology.