Paul,
Oops! You can register for free.
------------------------------------------------------------------
Thin-client showdown
End users will quash the idea that thin clients can replace their PCs.
By Geoffrey James
Network World, 1/19/98
Take a note: 1998 will be the year that sees the death of the notion that a network computer (NC) can replace a PC. When NCs were announced, they were positioned as the first real threat to the hegemony of the Wintel PC platform. Industry analysts at firms including Gartner Group, Inc., The Yankee Group and International Data Corp. estimated that an NC, or the later network PCs, would cost around $3,000 less per year to own and operate than a traditional PC, a sum that would get the interest of cost-conscious top managers.
This year, you'll find that thin clients - be they NCs, network PCs or Windows terminals - may find a niche as a replacement for dumb terminals, but one simple fact will deny them further glory: End users won't give up their PCs.
In a recent poll of 300 Network World readers, only 12% believed end users would agree to give up their PCs in return for a more stable and reliable thin client. Similarly, 72% of the respondents said they did not plan to deploy thin clients at all, be they NCs or network Pcs.
These findings were corroborated by a Zona Research, Inc. survey of 139 IT executives, 85% of whom said they had no intention of deploying thin clients. Based on the study, Zona concluded that "thin clients will largely have a user profile that reflects the users of traditional terminals.''
The results of both surveys are reflected in the lackluster commitment among hardware vendors for at least one type of thin client - the Microsoft Corp.-blessed NetPC implementation. In September, IBM decided there wasn't sufficient demand to justify building a separate line of NetPCs, a decision that quickly was echoed by Micron Electronics, Inc.
Other vendors say they remain committed to building a NetPC, but it doesn't appear the devices are on their front burners. For example, Gateway 2000, Inc. in May 1997 announced a NetPC but hasn't yet brought it to market, according to Jeff Hanson, a public relations manager at the firm. Even Digital Equipment Corp. - a company whose background in traditional time sharing would seem to make it an ideal candidate to embrace the NetPC - is taking a wait-and-see attitude and hasn't yet announced an implementation.
That approach is probably wise, says Rob Enderle, senior analyst at Giga Information Group, of Santa Clara, Calif. "The products haven't been selling particularly well,'' he says. He attributes the NetPC's failure to grab market share to a "screwed-up launch by Intel [Corp.], which didn't deliver the right message because it was afraid that [the NetPC] would threaten its existing architectures.''
Maybe so, but the survey results and other anecdotal evidence suggest there also is something fundamentally unappealing about the thin-client concept in general.
A clue to the real problem can be found, of all places, on Sun Microsystems, Inc.'s Web site. In its white papers section, Sun, a major NC backer, saw fit to publish a commentary addressing the "Danger of end-user reaction against NCs.'' It warns: "[End] users may view Network Computers as an attack on their independence, productivity and enfranchisement. If forced by management edict to switch to NCs, there are certain to be PC withdrawal symptoms in the form of complaints, intentional difficulties, unreasonable requirements and so forth.'' In other words, end users want PCs - they don't want Ncs.
"Even if the NC is better technology, it's probably not going to be accepted,'' says Tim Hoy, applications and systems manager at Omni Healthcare, a health maintenance organization that serves northern California. "The typical end user thinks of his desktop or laptop as 'his computer' - not as the company's computer. Who would willingly give up control of their own computer?''
Even companies that are planning to use NCs to replace dumb terminals don't plan to throw out their PCs. "We're planning to replace the X terminals in our call centers with NCs sometime in the future,'' says Sally Davenport, media relations spokesperson for Federal Express Co. "But we're certainly not going to take PCs away from people who need them to perform their jobs.''
Otto Folprecht, corporate IS manager at Tree Island Industries, a Vancouver-based manufacturer of steel parts, agrees the idea is of limited value. "I don't believe that managers are good candidates for NCs because they're going increasingly mobile," he says. "They want laptops or PCs with removable hard disks.''
A revolt at McDonald's
Despite all the talk about NCs, most implementations are in the prototype or pilot phase, so it's too soon to say for sure whether they'll live up to their promise. But if McDonald's Corp. is any indication, they surely won't.
McDonald's implemented an NC-like device several years ago, says George Vidinich, staff director for the fast-food chain. For select users at its Oak Brook, Ill., headquarters, the company installed desktop devices that, like today's NetPCs, were stripped of all application software. The devices did have their own C: drives, but these were to be used only for the DOS disk that was required to boot the machine and supply network connectivity, and as a cache to improve desktop and network performance. All Windows operating software, applications and data storage resources were server-based.
Users hated the new system, Vidinich says. They immediately began devising ways to get around what they saw as restrictions. For example, they figured out how to copy the operating system and key application executables onto their cache drives in order to get free of the network. This created multiple local versions of various pieces of server software, which added to the support burden - the opposite of what was intended. McDonald's finally decided to offer end users two desktop options: the NC and a more traditional PC LAN environment. There is now so little demand for the NCs that when McDonald's upgrades to Windows 95, it will kill the NC option altogether.
It could be argued that a modern NC implementation wouldn't have allowed the end users to subvert the system. However, what's interesting about the McDonald's experience is that end users thought it necessary to subvert the system in the first place. The reason, according to Vidinich, was system stability, one of the things NCs are supposed to improve. A major crisis occurred whenever the network, which had a relatively high uptime, went down. "People felt that it was stupid to have a computer on your desk that didn't work when the network isn't working,'' Vidinich says.
More risk and complexity
Traditional PCs connected to traditional PC networks allow individuals to keep their own data on their own machine and to keep data that's intended to be shared on the network. The network also provides a convenient place to back up personal data files. In such an environment, the failure of one individual's desktop device remains an annoyance rather than a disaster. While some data may be lost, the individual generally can find or borrow another PC and start working almost immediately. Similarly, if the network goes down, the inability to access shared data may be inconvenient, but most office workers will be able to switch to activities that don't require the network, using sneaker net or dial-up modems to share information.
With NCs, by contrast, every network failure is an unmitigated disaster because nothing meaningful can get done until the network is up and running. The solution, according to Greg Blatnik, a vice president at Zona Research, is "networks with very high uptime.'' How high? "One hundred percent uptime will be required,'' he says.
To make matters worse, networks that support thin clients will be more complex than current LAN designs because they'll have greater and different types of burdens, says Ellen Carney, worldwide director of network integration and support services in the Westborough, Mass., office of Dataquest, Inc., a market research firm. It's unclear whether shifting the support burden from the desktop to the network will result in a more stable system, she says. The desktops may become more stable when the network is running, but the network itself is likely to become less stable and more difficult to maintain. "Some companies may find that they've simply, in the words of the old commercial, traded a headache for an upset stomach,'' Carney says.
And even with 100% uptime, it's unclear whether end users would be happy. At McDonald's, for example, end users were accustomed to being able to predict whether they could get work done on time based on prior experience with the performance of their PCs. With the thin-client implementation, on the other hand, system performance slowed down during spike times - such as right before a deadline. Users found this unacceptable, not to mention frustrating. "We tried to take the 'P' out of the 'PC,' and that was a big mistake,'' Vidinich says.
Of course, system degradation during peak usage could be solved by adding massive amounts of server horsepower and network bandwidth that's unused most of the time - an investment that, if history is any guide, most companies are unwilling to make. And thin clients are supposed to save money rather than create demand for high-priced server hardware and redundant network hubs.
Cost savings from the thin-client implementation proved elusive at McDonald's, where the total cost of ownership went up rather than down, Vidinich says.
One reason was end users began purchasing more laptops in order to get work done untethered. This meant the IT group had to support the NC-like device and a laptop for many end users. "Unless you're willing to throw out every PC - including laptops - the NC simply becomes another platform to support, which only adds to your infrastructure costs,'' he says.
Sending the wrong message
Despite the fact that cost savings might prove elusive, some industry analysts - Blatnik for one - insist that NCs are a good idea. The important issue for Blatnik isn't system stability as much as it is system abuse. "I'm of the simple belief that when you're in the office, you shouldn't be doing 'personal' computing,'' he says. "Corporations are getting to the point where they won't tolerate abuses such as X-rated Web surfing, playing "Doom" and answering personal e-mail.''
But there's a downside to such top-down control of computing resources - the erosion of trust. "The NetPC, as currently implemented, is a safer PC from a management viewpoint, [but] it says some pretty ugly things,'' says Kim Brown, vice president and chief analyst at Dataquest's San Jose, Calif., office. "It says to employees that 'we [management] don't trust you.'''
Thin clients thus represent a way of thinking, a philosophy about the nature of employment and management, rather than being just a new technology.
Stripping the control of computing resources away from end users harkens back to traditional notions of top-down, industrial-age management.
That's exactly the sort of thinking management consultants say leads to business failure. However, the nail in the coffin of the "we can't trust end users'' argument is one of the major trends in today's business world - the extended workday. Salaried workers in America now work an average of 164 hours longer each year than they did in 1970, before computers came on the scene, according to Harvard University economist Juliet Schor. Mobile computing and telecommuting, rather than allowing professionals to work at home instead of at the office, have resulted in an increasing number of hours being spent doing tasks for the corporation.
It's one thing to demand that factory workers who punch a time clock stick to their appointed tasks. It's quite another to expect the typical 60-hour-per-week road warrior to work hundreds of hours of overtime without the ability to use the company's computers for a reasonable amount of personal business and, yes, even recreation.
That's why end users simply aren't going to buy into the thin-client story. Having already paid the price of being "empowered'' by the PC, they're not likely to give up the benefits they receive from "owning'' their computing resources. Even NC proponents such as Blatnik agree this is the case.
"We've given people the idea that freedom and automation go hand in hand,'' he says. "If you give the impression that you're taking that away, you could end up with Mutiny on the Bounty."
--------------------------------------------------------------------------------
stephen |
