Click on our Sponsors to help Support SunWorld

Finding value in NCs

How much money can you really save? What should you consider before implementation? We give you the figures and the deciding factors

By Dave Kosiur

June  1997
[Next story]
[Table of Contents]
Subscribe to SunWorld, it's free!

While not a panacea to all personal computing needs, NCs do have a place in the enterprise. Offering savings in support costs and, to a lesser extent, capital costs, network computers can provide many users with increased functionality and improved performance. (2,000 words)

Mail this
article to
a friend
The concept of NCs, or network computers, as evangelized by Larry Ellison and Scott McNealy, among others, has been around for just over a year. It has forced network and IT managers to reconsider what it costs to maintain their corporation's personal computing resources and upgrade paths for much of the equipment accessing legacy systems. NCs have also crystallized or at least re-emphasized arguments of users' freedom of choice and PC customization, centralized control of computing resources along with many other concepts revolving around distributed computing.

While NCs are not a panacea to all personal computing needs in a corporation, they do have a place in today's IT structure. Let's take a look at some of the advantages and disadvantages of NCs and where they fit in enterprise networks.

What's an NC?
As envisioned by some, the term "network computer" can mean almost any computing hardware that's smarter than a dumb terminal, but not as complicated (or as expensive) as a PC. Some aim NCs at the consumer market as well as the business market, but for the purposes of our discussion, we'll limit comments to the business market only.

Individualized computing forms a spectrum of choices, ranging from dumb terminals like VT100 and IBM3270 terminals connected to minis, mid-size or mainframe computers all the way to PCs and workstations that can function either as standalone computers or on a network. The following table shows how NCs and other devices fit into this spectrum.

Computing Platforms
Dumb terminal Little, if any, internal intelligence; data resides on and applications execute on host computer $500 - $1000+ Low cost of ownership
NC Runs software over network $500 - $1000 Low cost of ownership
NetPC Runs software locally or over network Greater than $1000 Low cost of ownership; maintains PC format
Low-cost PC Runs software locally Starts below $1000 Low initial cost; maintains PC format
Traditional PC/workstation Runs software locally or over net, most power at desktop $2000+ Most power; greatest customizability

There's been an attempt to create standard definitions for the NC, but so far the Network Computer Reference Profile first suggested by Oracle does not impose many limitations on implementations, making the definition too open-ended. Being hardware-neutral, the Network Computer Reference Profile concentrates on software and concepts. For example, it specifies support for important networking protocols, such as TCP/IP, NFS, FTP, Telnet, SNMP, BOOTP, and DHCP, JPEG, GIF, WAV, and AU for multimedia, HTML, HTTP, and Java for the Web, as well as 3 protocols, SMTP, IMAP4 and POP3, for e-mail.


Is it really a question of cost?
When the term "network computer" was first introduced, some of its primary advantages were the unit's low cost and a lower cost of administration. Granted, NCs can cost less than fully decked-out PCs and workstations, but they're often only marginally less expensive than low-cost PCs and should see price competition from NetPCs as well, whenever they become available.

More intriguing are the costs of administration and support. Everyone seems to agree that it costs less to administer a network computer, but no one can agree on just how much money NCs save, partly because everyone has a different idea of how much a PC (and its support) costs. Aiming at a networked PC running Windows 95, we've got the following estimates:

Gartner claims a cost of $6776 for supporting an NC with server; the biggest differences between PC and NC costs in this estimate are capital ($1896 vs. $2532), administration ($771 vs. $1497) and user responsibilities, which includes training and peer support ($2611 vs. $4052).

While the numbers may differ radically, the general consensus seems to be that NCs lead to a cost savings over PCs of at least 30 percent and may rise as high as 45 percent.

As a real-life example, consider what Don Resh, senior vice president at Retired Persons Services Inc. in Arlington, VA, estimates for NC-related savings. By introducing 1000 new NCs, RPS spent $1.9 million for the thin clients and new servers, as compared to an estimated cost of $2.45 million if PCs had been purchased. Similarly, the estimates for a five-year maintenance and support program total $2.5 million for 1000 NCs and $35 million for the same number of PCs. (Yes, those numbers are correct!) Resh estimates more than a factor of ten difference for supporting PCs.

But NCs aren't always going to be purchased in place of networked PCs. In fact, in many situations, it's highly likely that NCs will replace dumb terminals in large numbers. Many companies, such as RPS, Sears, and Federal Express, are already starting that movement. In such cases, NCs may well cost as much as 25 percent more than the terminals. Of course, NCs offer more capabilities than the dumb terminals, so you are getting more for your money.

But neither capital costs nor support costs should be the only deciding factor in selecting NCs for use on a corporate network. Other factors, including your company's information infrastructure, organizational politics, user requirements, and software development skills, are equally influential.

If your company is already using centralized databases and application servers, then NCs fit well into such an infrastructure. This may also be well suited for IS departments that have grown up with terminal-based access to mainframe data and applications. In fact, the general concept of network computers is very similar to that of terminal networks, with the significant difference that NCs have sufficient computing power to participate as active members in client/server architectures (which terminals cannot).

For users more accustomed to distributed computing or PCs that can function as standalone or networked computers, the idea of increased dependence on centralized servers administered by the IS department may well seem like a step backwards. But there is a growing trend among corporations to consolidate data and application servers -- it's not a matter of control as much as an attempt to increase the accountability for resources and justification for existing computing resources.

User requirements have to be considered when deploying NCs. It's highly unlikely that NCs are suitable for many power users. On the other hand, many employees using terminals or those using only one or two applications on a PC are likely candidates for NCs. In the case of terminal users, NCs offer the same functionality (IBM3270, IBM5250, and VT100 terminal types, for example) as terminals, but provide more options for future expansion, particularly as departments shift to client/server systems using Java. It's important to consider the "demographics" of your users, leaving early adopters and power users to their workstations and networked PCs while using NCs for those on your staff who are just interested in getting their jobs done, without fiddling with a PC or similar system.

Lastly, you should consider the current and planned software development skills of your company. If you aren't already using Java, but plan to, then NCs shouldn't pose a problem. On the other hand, if you're committed to developing applications in-house but don't use, or plan to use, Java, you should think hard about how you'll support NCs. Yes, they can use stock Java apps and can interface with Microsoft Windows applications, but much of their power lies in Java.

Is an NC limiting? For whom?
Power users are likely to complain that NCs are too limited in functionality and that many of their required programs aren't available for them.

First, NCs aren't designed for power users, so they shouldn't be targeted at them. Let them work with a workstation or networked PC -- futzing with little details on their "personal" machine and developing new applications that they might otherwise not be able to do on an NC.

But the complaint that applications aren't available is a hollow-sounding one. Sure, if we restrict ourselves to Java-based equivalents of software available for Windows, Unix, and other operating systems that's true. The Java-based apps are only now starting to increase in availability. But there are other ways to use an NC to use the same Windows apps as a PC user, for example, using Citrix Systems' WinFrame, Exodus's NTerprise, or Insignia's NTrigue.

Employees accustomed to using dumb terminals most likely won't feel limited by using NCs. After all, most of the current batch of NCs offer some type of terminal emulation, particularly focusing on IBM 3270 and 5250 as well as Digital VT100 types. What the NCs will offer former users of dumb terminals are more possibilities to use newer networking technologies and applications, especially as corporations move to intranets.

NCs may not pose many limitations on functionality, but they do have certain network-related restrictions. While NCs aren't for Java alone, they are focused on TCP/IP, not IPX or NetBIOS, so you'll have to consider what networks your clients currently use before switching. Of course, with the popularity of intranets and TCP/IP networks in corporate environments, NCs are a natural fit.

Also, the focus of NCs thus far has been on execution and display, not printing. No NC standards have been publicly proposed for handling printing from NCs, so network printing is probably the best way to start out, waiting for vendors to determine how they'll support local printing.

Can you manage?
Network computers are supposed to make it easier to manage the software and data that everyone uses on a corporate network. To a point, that's true; but there's a paradigm shift in software distribution associated with NCs. Since most NCs do not have local storage, NCs are expected to download the applications they require from one or more servers on the network. This procedure makes it easier for IT managers to create new client software and then distribute the upgrades from a centralized location. With technologies like Marimba's Castanet, for example, gone are the worries that someone will have an outdated client.

The NC Reference Profile includes support for SNMP so that some of the NC's resources can be managed over the network; but SNMP wasn't originally designed for managing personalized computing resources. A better system is required if you're going to monitor NCs and, if you're really a control freak, figure out what individual users are running on their NCs.

Each NC vendor has gone its own way for NC administration and management. It's possible that an SNMP MIB will be developed for NCs, which would include more system and user administration functions than typically found in SNMP. But this hasn't happened yet.

Also bear in mind that the lack of local storage doesn't prevent NC users from customizing their computers. NC users will require storage space on a networked server, and they may choose to keep special applets or other applications there. It may prove easier for IS to control the amount of disk storage by monitoring servers, but most managers are going to have better things to do than routinely check the contents of each user's server directories.

Deciding factors
When you're planning for NCs in your corporation, consider that they will work best in specific situations:

Also, be aware that different users will have different views towards NCs -- from the viewpoint of a PC or workstation user, NCs may well represent a question of control; from the viewpoint of a dumb terminal user, it may be more of a question of changing from a character-based system to a GUI.

Click on our Sponsors to help Support SunWorld


About the author
Dave Kosiur is a free-lance writer and networking consultant based in Reston, VA. His latest book, Understanding Electronic Commerce (Microsoft Press), is due out this month. Reach Dave at

What did you think of this article?
-Very worth reading
-Worth reading
-Not worth reading
-Too long
-Just right
-Too short
-Too technical
-Just right
-Not technical enough

[Table of Contents]
Subscribe to SunWorld, it's free!
[Next story]
Sun's Site

[(c) Copyright  Web Publishing Inc., and IDG Communication company]

If you have technical problems with this magazine, contact

Last modified: