This article was published in CAUSE/EFFECT journal, Volume 21 Number 3 1998. The copyright is shared by EDUCAUSE and the author. See http://www.educause.edu/copyright for additional copyright information.
Considering Thin Client Computing for Higher Education
by Mark SheehanDesktop computer technology has been following a well-defined growth path for the past twenty years: personal computers (PCs) have become more powerful, their user interfaces have become more graphical, and they�ve become more widely interconnected with others of their kind.
It�s taken a while for applications technology to catch up with the potential of the desktop computer. In the last decade, client/server applications allowed us to stop using our $2,000 desktop computers as $200 terminals. Client/server applications and PC technologies have been developing in tandem, each feeding the other�s momentum. Desktop computers have become bigger (in terms of storage and memory) and faster. In response, client applications have become more fully featured.
This expansion of features in both PCs and applications has not come without a price, however, and while the price may not be slowing the progress of PC and application development, it is causing the direction of technological development to change. We are now at a crossroads in the evolution of desktop computer technology, and this is a good time to stop and take a look around.
Agents of change
Several cost factors are changing the directions in which desktop technologies will grow in the next century. Purchase price has always been an issue. While the cost per unit of processor speed, memory, and disk has been shrinking steadily for twenty years, the need to increase those capacities has been growing at about the same rate. In 1982 an adequate PC cost $2,000. It costs about the same now. The difference is in what we consider adequate.
The life span of a PC is a related issue. It�s bad enough to have to spend $2,000 for a PC; it�s worse to have to replace it every two or three years! But the pace of technological change often requires that. When we used old PCs as terminal emulators to access legacy, mainframe-based software this wasn�t such an issue. A ten-year-old PC is as good a terminal emulator as a new PC. But now, as many of us adopt year-2000 compliant, client/server administrative software and use multimedia Web sites for our communications needs, the PC client software we use is more complex and more of us--sometimes hundreds more of us at once--need up-to-date PCs on our desks.
Finally, we�re beginning to realize the extent to which support costs distort the cost-benefit balance sheet of desktop computing. McClure, Smith, and Sitko provide an excellent overview of these issues behind these costs, which include growing numbers of support consumers, increasingly sophisticated applications of computer systems, lack of standardization in hardware and software, and archaic funding models for support.1 Industry calculations of the total cost of ownership (TCO) of a PC now take into account support costs and the total figures are staggering. A 1997 Gartner Group article estimates the TCO for a networked Windows 95 PC at about $8,000 per year.2 Network costs add about $2,000 per year.
Taking the load off
Concern over TCO has prompted, or at least accelerated, the development of several new desktop solutions. They are generically referred to as thin clients or network-centric computers. These umbrella terms embrace two distinct types of desktop device: the network computer and the Windows terminal. A third type, the network PC, is just a scaled-down PC and has had no real market success. We can safely forget about it.
Network computers
Network computers (NCs) are powerful microcomputers with lots of memory but no disk storage: no floppy disk, no hard disk.3 They are connected to a network, and through the network to a server. They are utterly dependent on the server computer for their operating system software, their applications, and their file storage. Each time the NC is booted it pulls down its operating system anew. Each time it launches an application, that software is downloaded to the NC afresh. Applications run on the desktop, on the NC�s processor using the NC�s memory.
NCs run two sorts of applications: those written for the NC�s processor ("native" applications) and Java applications. The range of native applications for NCs is very narrow at present. The most pervasive are Web browsers, though Java-based browsers are also available. Other Java-based applications include Lotus�s E-Suite (a multi-function office productivity package) and a number of Oracle data management applications. Several companies (e.g., Corel) have suites of Java-based applications under development, though as of this writing those projects show signs of having stalled out.
Windows terminals
The Windows terminal (WT) is harder to describe in physical terms because it can be practically any kind of desktop computer. Being a WT isn�t so much what a thing is as what it does. An old 386-based PC can be a WT, as can a new Pentium or a Mac or a Unix workstation. Several companies even make dedicated WTs, which serve no other function. (Incidentally, many NCs are also capable of acting as WTs.)
A Windows terminal contains its own operating system, either on its hard disk or on a chip. Unlike the NC it doesn�t download its operating system over the network. It does rely on the network for its applications, but unlike an NC the WT doesn�t download the application code and run it internally. Instead, the WT�s applications run in the server�s processor using the server�s memory. The WT�s data files are customarily stored on the server. The WT has to do only two things: send keyboard and mouse input to the server, and paint the screen with output from the server.
WTs can access almost all 32-bit Windows applications, including the full, "thick" Windows versions of the PC client software for the major commercial suites of higher education administrative software. At considerable expense to server performance, WTs can be given access to many older 16-bit Windows applications.
Can thin clients help?
To different degrees, thin clients address our three major issues: purchase cost, life span, and support costs. Their effect on total cost of ownership is still being calculated, but seems promising.
Purchase cost
The most complex issue is purchase cost, precisely because thin clients are network centric: they can�t exist alone; they require networks and network resources. Table 1 compares the purchase costs of PCs with those for three types of thin clients. Network costs are not included because they are about the same for each type of installation. Network architecture considerations are important, nonetheless; if you have a network in place it may need some serious reworking before it supports thin clients optimally.
Table 1: Purchase cost comparison: PCs versus thin clients
Component
Device TypeWorkstations Monitors Server
(estimated)Licenses Total Cost 50 PCs $100,000
(50@$2,000)Included None Included $100,000 50 NCs $37,500
(50@$750)$12,500
(50@$250)$7,500 Included $57,500 50 Dedicated WTs $37,500
(50@$750)Included $22,000 $3,000
(NT Terminal Server)$62,500 50 Legacy PC's as WTs Prior Investment Included $22,000 $16,000
(NT Terminal Server+MetaFrame)$38,000 The need for NC and WT servers, of course, implies the need for server administrators. Most higher education computing environments already have LAN server administrators who could be developed into NC or WT server administrators. However, WT server administration, in particular, is a much bigger job than LAN administration, and building a WT environment may require adding server administration staff. It�s worth adding that NC servers have a reputation, verified by our own experience at Montana State University, for being very difficult to install.4
Life span
Windows terminals help avoid some of the rapid obsolescence problems we experience with PCs. The fact that a ten-year-old PC can be used as a WT is evidence of this. In fact, if the old PC has a high-resolution video card in it, the user will perceive very little difference between the PC and a brand new, dedicated WT. WTs weather software upgrades well, too. If an application grows to require more memory and disk, the server administrator makes those upgrades once, on the server, rather than upgrading every workstation. Changes will doubtless occur in the software that lets a legacy PC act as a WT (Citrix MetaFrame), with minor software maintenance implications. But such minor changes should not require the periodic hardware upgrades that current mainstream PC software does.
In the NC world, as users� expectations grow, the software market will respond with bigger, more complex Java applications. To keep up, NCs, which do much more with their on-board processors than NTs do, will probably require periodic hardware upgrades. For example, many recent NC buyers, wishing to adopt the Lotus E-Suite applications, found they needed to add more memory to their NCs. That is probably only the beginning of a trend.
Support costs
NC and WT manufacturers have capitalized on recent concerns about the total cost of ownership of desktop systems. For decades, they argue, IT administrators have looked too narrowly at hardware, software, and network purchase and installation costs, have given lip service to training and support costs, but have ignored or denied the so-called hidden costs of PC ownership. Most of these hidden costs are related to productivity and the amount of time information workers spend customizing, souping up, or otherwise "futzing" with their computers.
A network of NCs or dedicated WTs, properly run, can eliminate many productivity sinks. The server administrator determines with absolute authority which versions of the operating system and application software are used by the thin clients on the network. Files are backed up regularly, under central control. With NCs and dedicated WTs, users have little ability to customize and needlessly complicate their computing environments (by bringing in software from outside the enterprise, for example).
When existing desktop computers are used as WTs, however, the server administrator may have complete control over centrally served applications, but no control over desktop configuration issues--likely the biggest support cost sink. In some scenarios, that vulnerability will counteract the favorable purchase price of legacy-PC WTs (see Table 1).
Controlling the computing environment doesn�t ensure good productivity, of course; human nature is (so far) not susceptible to technological controls. But a controlled computing environment can simplify training and support requirements and can eliminate many temptations to unproductive diversion of effort.
Overall TCO
Only long-term tracking of NC and WT use can bring us accurate TCO figures, but a number of industry groups have made predictions. In spring 1996, Zona Research predicted a five-year, total cost of ownership savings of 57 percent for Wyse Windows terminals versus PCs.5 In summer 1997, Microsoft suggested that savings of 46 percent in TCO were possible in the same scenario.6 In late 1997, Gartner Group predicted a 22 percent savings for NCs versus PCs.7
A return to the bad old days?
The new network computer and Windows terminal environments blend the best of the PC world--the high degree of network connectivity, the graphics, the user friendly software--with the high-productivity focus of the centrally controlled environment. At the same time, NCs and many WTs have lost one of the initial benefits and attractions of the desktop computer: autonomy.
Institutions that are about to move from a mainframe-terminal environment into a new client/server environment will want to consider the potential benefits of NCs and WTs. They make excellent upgrades for heads-down users of "greenscreen" terminals, and they can be a cost-saving alternative for library catalog access and for staff in higher education customer service environments such as business office, financial aid, and advising.
However, in offices, classrooms, and laboratories--where PCs are equipped with a variety of software tools and their users frequently push the limitations of the machines--NC or WT environments may prove much too confining and may actually decrease productivity. Research faculty will probably never embrace thin clients, for example, nor will most other higher education knowledge workers. The flexibility of the desktop computer environment, expensive as it is to buy and support, brings such people too many important advantages.
NCs and WTs aren�t for everyone. But they represent a change, or at least a branching, in the evolutionary direction desktop computing has taken in the last twenty years. Viewed conservatively, in relation to the old, well-established direction, thin clients may appear to be recidivist: a return to the bad old days of central control and authority. Viewed more broadly, though, thin clients appear to be defining a direction of their own. Unlike PCs, thin clients won�t pop up everywhere, but where they make sense they can simplify computing environments and save money.
Sidebar
Internet Resources for Thin Clients|
and Related ProductsNetwork Computers
Sun Microsystems JavaStations (http://www.sun.com/javasystems/krups/)
IBM Network Station (http://www.pc.ibm.com/networkstation/products/)
NCD Explora (http://www.ncd.com/pexp/pexp.html)
Neoware NeoStations (http://www.neoware.com/neostation_info.html)
Acorn Corporate NC (http://www.acorn.com/acorn/products/nc/ corpnc.html)Windows Terminals
Wyse Winterm (http://www.wyse.com/winterm/)
Tektronix ThinStream
(http://www.tek.com/VND/Products/ Network_Computers/18w-8013-2.html)
Boundless Viewpoint (http://www.boundless.com/network/)
NCD Thinstar (http://www.ncd.com/thinstar/thinstar.html)
Neoware @workstationServer Products
Microsoft Windows NT 4.0, Terminal Server Edition
(http://www.microsoft.com/ntserver/basics/TerminalServer/default.asp)
Citrix WinFrame (for Microsoft Windows NT 3.X servers)
(http://www.citrix.com/products/winframe.asp)
Citrix MetaFrame (for WinFrame and for Microsoft Windows NT 4.0, Terminal Server Edition)
(http://www.citrix.com/products/ metaframe.asp)General Information
NC World online magazine no longer publishes new material as of July 1998, but has valuable archives still available (http://www.ncworldmag.com).
Endnotes
1 Polley A. McClure, John W. Smith, and Toby D. Sitko, The Crisis in Information Technology Support: Has Our Current Model Reached Its Limit?, CAUSE Professional Paper #16 (Boulder, Colo.: CAUSE, 1997). This paper is available online at (http://www.educause.edu/ir/library/html/pub3016/16index.html).
2 Gartner Group, "TCO: New Technologies, New Benchmarks," Gartner Group�s Managing Distributed Computing Research Note TCO-252, 5 December 1997.
3 "Network Computer™ Reference Profile," The NC Reference Profile Web home page (http://www.nc.ihost.com/nc_ref_profile.html).
4 Nicholas Petrely, "The Network Computer�s Dirty Little Secret," NC World online magazine, March 1998 (http://www.ncworldmag.com/ncworld/ncw-03-1998/ncw-03-straypackets.html).
5 Zona Research, Inc., Desktop Clients: A Cost of Ownership Study, Spring 1996 (http://www.wyse.com/solution/tco/intro.htm).
6 "Well Managed Windows-Based Solutions Offer the Lowest TCO," Microsoft Market Bulletin, February 1998 (http://www.microsoft.com/windows/platform/info/gartnertco.htm).
7 Gartner Group, "TCO: New Technologies, New Benchmarks."
Mark Sheehan ([email protected]) is director of the Montana State University Information Technology Center and served on the (former) CAUSE Current Issues Committee.
