The spin cycle, 09/27/99

By Paul Desmond
Network World, 09/27/99

It takes time for an emerging technology to generate a full-fledged buzz, a real "can't go the day without hearing something about this" kind of thing. Then, just when honest-to-goodness products make their way into enterprise networks, the buzz fades. That's the spin cycle for you. Throughout this issue, we 've dug into various technologies spinning through this cycle; here we offer a brief look at 10 more - corporate portals, Gigabit over copper, super-fast routers, wave division multiplexing, digital subscriber line services, personal digital assistants, Web caching, Windows 2000, e-mail management and public-key infrastructure.

Corporate portals
DSL
E-mail
Gigabit over copper
PDAs

PKI
Super-fast routers
W2K
WDM
Web caching

Portal potential

Portal talk
Network World columnists Tom Nolle, Dave Kearns and James Kobielus deflate portal hype (6 min. in G2 RealAudio).

John Gregory, a marketing specialist for the U.S. Postal Service, is sold on the idea of corporate portals. When he came to the agency in November 1998, he inherited a custom portal project. The portal was the latest in a line of projects aimed at giving some 3,000 internal sales and marketing personnel easier information access.

Originally, users got at the data - information on competitive companies and products or potential customers, for example - via a traditional client/server network. Then the Postal Service went to an intranet and saw usage double. The portal, launched the first week of July, should continue that growing usage trend, Gregory says.

"Ease of use, more creative design to make the important stuff obvious, and more-powerful search tools will drive the growth," he notes.

Through a portal, each user gets better links to data - internal and external - needed to get the job done. And network managers who have portal software at their disposal get tools that make it easier for them to tie databases, application servers and other resources into Web servers.

Guy Creese, senior analyst at Aberdeen Group in Boston, says a portal should be able to handle structured data, such as reports and tables, as well as unstructured data such as documents. A portal also needs to be scalable, meaning it must have the ability to keep up as demand increases, such as by running on multiple servers simultaneously.

Whether a given portal will be able to do all that depends largely on the kind of vendor offering the portal. Vendors whose forte lie in document management, enterprise resource planning, search engines or any number of other disciplines are adding Web front ends and some rather limited integration tools and declaring themselves portal vendors.

To find a "pure" portal vendor, turn to a start-up such as Epicentric, Plumtree Software or Portera Systems. These companies have more incentive to provide tools for helping you integrate all kinds of data into your portal, as Gregory discovered.

He has been beta-testing Epicentric 's product, which he says makes the process of integrating internal and external resources "less daunting" than doing it yourself. And on the external side, Epicentric helps negotiate with content providers.

It 's early in the game yet, but analysts expect in a year or two portals will play a key role in enterprise nets. Kevin Warbach, managing editor of the newsletter "Release 1.0" from EdVenture Holdings, is especially bullish.

"Long term, the corporate portal becomes the desktop and potentially the operating system for the enterprise," he says.

Whoa, now. First things first.

Related links:

Procurement portals for IT
Online marketplaces offer one-stop shopping for network hardware and services.
Network World, 8/23/99.

E-commerce portals open for business
Companies old and new are tooling up to tie businesses, and their many partners, together.
Network World, 5/3/99.

Microsoft, Nextel need to rethink wireless portal strategy
Network World, 6/7/99.

Gigabit over seaweed

It wasn 't so long ago that the very idea of Gigabit Ethernet was enough to cause a buzz. Now Gigabit Ethernet is practically ho-hum; everybody 's doing it. The only challenge is getting it to work over cheap, crummy old wire.

OK, maybe Category 5 copper isn 't exactly as bad as all that, but it 's definitely less expensive than fiber. And that 's attracting users to the idea of gigabit over copper.

So as you can imagine, all the mainstay Gigabit Ethernet vendors are fairly champing at the bit to get products out the door, if they haven 't already. Some vendors experienced delays because their chip suppliers, mainly Broadcom and Level One Communications, took longer than expected to deliver the chips that make all this possible.

But you can expect a steady trickle of gigabit-over-copper products to be flowing by year-end. It won 't be until that trickle becomes a stream, however, that you 'll see the price benefits that the technology promises.

That's because pricing from the chipmakers isn't as aggressive as some switch vendors had hoped. It was widely expected that gigabit-over-copper prices would be roughly half that of fiber versions. But when the chips started shipping in July, they cost about $75. Meanwhile, gigabit chips for fiber cost $60 to $70, or $50 for very high volumes, says Drusie Demopoulos, vice president of marketing at Foundry Networks in Sunnyvale, Calif. Gigabit copper chips will probably drop to $45 to $50 by the middle of next year, when they'll be shipping in volume, she adds.

Nonetheless, Demopoulos says, users are pressuring switch vendors to keep gigabit copper prices low. For its part, Foundry Networks announced last month that its copper-based FastIron switches will cost $650 per port, or roughly one-third the price of its gigabit fiber switches.

Even with low prices, the gigabit-over-copper market won't explode any time soon. Most end users simply don't yet need that kind of bandwidth to their desktops. There's also a distance limitation issue in that the technology generally works only up to about 100 meters, whereas fiber can stretch for kilometers.

Nonetheless, gigabit over copper will be a good, inexpensive way to link switches to servers and to each other.

By the way, the IEEE has formed a working group to investigate 10G bit/sec Ethernet. Can gigabit over seaweed be far behind?

Related links:

Net Resources: Gigabit Ethernet
Includes an audio primer, articles and an overview of the technology.

Son of Gigabit Ethernet
10-Gigabit Ethernet is coming soon to a LAN (and MAN) near you.
Network World, 8/9/99.

Gigabit maturity
Network World Fusion, 7/26/99.

Ridiculously fast routers

If gigabit speed isn't fast enough for you, don't worry - plenty of vendors are now shipping terabit routers. Of course, if you think you can squeeze by with gigabit routers for a while, you'll have even more from which to choose.

Vendors in the gigabit/terabit router market include start-ups and stalwarts, from Avici Systems, Juniper Networks and Pluris to Cisco, Nortel and 3Com. And then there are the start-ups that have been acquired by stalwarts, namely Argon (Siemens), NetCore (Tellabs) and Nexabit (Lucent).

No matter the provider, one thing is clear: Cust-omers are overwhelmingly carriers and ISPs - a fact that no one expects to change much anytime soon.

Rosemary Cochran, principal at Vertical Systems Group, a research firm in Dedham, Mass., says these core IP switch/routers, as her company calls them, serve two main purposes for service providers: They provide more raw throughput and, in the longer term, bring in a platform capable of carrying voice and data over IP.

As for enterprise users, only those that are already using OC-12 connections will need to consider a terabit-speed router, Cochran says.

Enterprise users are more interested in gigabit routers, says Rick Forberg, director of marketing for LAN systems at 3Com. He reckons companies with more than 2,000 users in the same campus net would have a serious interest in gigabit-speed Layer 3 switch/routers. Of those, maybe 25% to 35% have near-term deployment plans, he says.

So what's in it for the average user in all this terabit talk? Essentially, you get to keep using the Internet without experiencing horrific performance problems.

"If we don't keep cranking up the throughput of the switching points, the congestion's going to get worse and worse," says Bob Bellman, principal at Brooktrail Research in Natick, Mass. "There's not much more to be said about it than that."

What about price decreases, given the greater economies of scale these terabit behemoths bring carriers and ISPs? Not so long as the demand is there, Bellman says.

And what of advanced QoS features that the high-speed router vendors promise? Those are being worked on, but right now interoperability problems abound, Cochran says. "There's no core IP router that's standard to the point where it's totally compatible [with other vendors' equipment]."

In short, unless you're one of the biggest fish in the enterprise sea, focus your attention on some other buzz topic and let the service providers worry about this one.

Related links:

Getting more bang from your routers
Network World, 7/5/99.

Is Lucent wise to the enterprise?
Network World, 7/5/99.

Lucent to snap up terabit router vendor
Network World, 6/25/99.

A terabit secret no more
Network World, 5/19/99.

Are megarouters coming to your net?
Network World, 5/17/99.

Fast router feast
Network World, 11/2/98.

Dopey predictions of free bandwidth

If you had a nickel for every time you heard some pundit predict that bandwidth will cease to be an issue in the next year or two, odds are you'd have at least a couple bucks and the goods on some otherwise fairly reasonable industry watchers.

Sure, wave division multiplexing (WDM) and the closely related "dense WDM" (DWDM) are fairly amazing technologies that can help you increase the capacity of a single fiber-optic line by at least fourfold. But if you think for even a fleeting second that will translate into "free bandwidth" - or anything close to it - then you've got to spend a little more time trying to understand this whole capitalism thing.

Definitions of the technologies vary somewhat, and the terms are often used interchangeably. But if you get four waves out of a fiber, that's basically WDM, and if you split a signal into eight or more waves, that's DWDM, says Richard Barry, architecture director at optical network switch-maker Sycamore Networks.

As is the case with gigabit and terabit routers, carriers will reap benefits from WDM and, to a greater extent DWDM, which is the technology of choice in long-haul nets. Of course, there are also applications for WDM in fiber-based LAN and campus environments. But in many cases, analysts say, it will be less expensive for companies to simply pull more fiber through existing conduit, or light dark fiber that's already installed, than to buy WDM equipment.

The metropolitan area could be more interesting, however, when coupled with the increasing number of competitive local exchange carriers offering fiber - dark or otherwise. Even on the low end of the WDM scale, the technology would let companies extend campus LANs across a metropolitan area with little or no loss in speed.

Here again, though, the number of companies that need the capacity WDM can provide is relatively small. "The masses are more than content if they can beat T-3 at this point," says 3Com's Forberg. Demand will someday catch up with supply, but for now WDM/DWDM is another technology of interest mainly to service providers.

Related links:

Making bandwidth free
Network World 200, 4/26/99.

Cisco spends $7.4 billion on optical push
Network World Fusion, 8/26/99.

DWDM comes in like a lion, out like a lambda
Nolle's view. Network World, 6/14/99.

DSL will dominate

When it comes to supporting remote office workers and telecommuters, it looks like digital subscriber line (DSL) services will rule the roost. Cable modem services are still seen as more of a consumer option than a business one - and for good reason.

Cable networks are fraught with security and bandwidth-contention problems stemming from the fact that cable is a shared medium, says Lisa Pierce, director of global telecommunications services for Giga Information Group in Cambridge, Mass. If you and your neighbor have cable modem service, it's highly likely you're sharing the same bandwidth.

DSL, on the other hand, gives you guaranteed bandwidth on a line that's all your own - at least to the local carrier's point of presence.

Among the many flavors of DSL, Pierce favors symmetric over asymmetric because the former supports bidirectional high-speed links that operate to around 768K bit/sec. ADSL, on the other hand, supports a much faster downlink than uplink. Workers in remote offices will often be creating large documents and will need the faster uplink pipes, she says. That may become a moot point, however, as carriers continue to offer ever-faster versions of ADSL, with upstream links up to 896K bit/sec.

This, of course, is assuming you can get DSL at all. While in large metropolitan areas you're likely to have your choice of cable, ISDN and probably a number of DSL providers, the story is far different in more remote areas - or even not-so-remote locations that just happen to sit too far from a central office to support DSL.

There are other issues, including standardization, especially for SDSL, and the ability to support voice, which providers have been slow to do. Many providers are also relatively small companies that may not remain economically viable on their own.

Still, there's just too much momentum behind DSL for it to fail. Dell'Oro reports that in the first quarter of this year, the DSL equipment market grew 51%, while the market for head-end and client-side cable equipment grew 14%, both vs. the previous quarter. That's telling given the size of each market is comparable - first quarter 1999 revenue of $155.3 million for cable and $150.4 million for DSL.

PDAs coming of age

PDA talk
Network World columnists Tom Nolle, Dave Kearns and James Kobielus deflate PDA hype (6 min. in G2 RealAudio).

About 80% of the Palm organizers sold by 3Com's Palm Computing unit are used in work environments and 50% of them are ultimately paid for by a corporate account, says Chuck Yort, senior director of enterprise sales for Palm. And some 1.9 million smart handheld devices will be sold this year, according to International Data Corp., a market research firm in Framingham, Mass.

Still think you don't need to worry about these sorts of personal digital assistants (PDA)?

At the same time that sales are booming, new applications are emerging that will have PDAs pounding on more and more enterprise nets, with the blessing of central IS.

For example, Palms figure heavily into a data entry and collection system that American Medical Response is building for San Mateo County in California, says Eric Gee, project manager for AMR, a nationwide ambulance provider based in Aurora, Colo.

Paramedics will use Palm units to complete patient care reports on each emergency. At the end of a shift, each paramedic will upload all the reports to a central database.

Emergency 911 dispatchers, ambulance providers and hospitals will be feeding data on the cases to that database as well. Tapping into the database, AMR will be able to research which neighborhoods have a higher incidence of cardiac arrest, for example, and make sure local paramedics are equipped accordingly. The database also should help in training paramedics by enabling comparisons of the original problem assessment with the ultimate diagnosis and outcome.

"The Palm is the core of everything," Gee says. Currently, laptops and similar equipment cost $4,000 to $8,000 apiece, making their widespread use cost-prohibitive. "We can put a Palm in everybody's hand for $300."

Diana Hwang, research manager at IDC, expects PDAs to find their way into the enterprise, but she also expects them to bring management and security problems. "If you've got sensitive company data on one of these devices, what happens if you lose the thing?" she asks.

Vendors are dealing with these issues. 3Com and Computer Associates in July announced software that would enable CA's Unicenter TNG to distribute software to Palm devices and pledged to develop additional Unicenter asset and inventory management and remote control tools. For security, 3Com has encryption wares that will protect corporate data, Yort says.

Despite the drawbacks, users expect big things of their Palms. IDC conducted a survey on user perceptions regarding handheld computing devices. More than three-quarters of the 300 plus users who said they used the devices mainly to increase productivity said they were realizing gains of between one and four hours per day. And that survey was conducted last year.

Related links:

Putting wireless network power in the palm
Mark Gibbs on the Palm VII. Network World, 9/6/99.

3Com spins off Palm Computing division
Network World, 9/20/99.

Psion cites poor sales in first half results
Network World Fusion, 9/17/99.

Cache me out

Web caching is a tough technology to handicap. Like big huge routers and WDM/ DWDM, there's no question Web caching is great for ISPs. It's less clear whether enterprise users will find the technology useful.

Web caches store copies of oft-accessed Web pages, enabling end users to access the pages from the nearby cache instead of the faraway Internet. This reduces response time and conserves wide-area bandwidth.

Peter Christy, managing partner for research at the Internet Research Group, a consulting firm in Los Altos, Calif., expects the market for Web caching products will grow from $88 million in 1998 revenue to $1.6 billion in 2002.

How much of that money will come from enterprise customers is the big question. "So far, the enterprise hasn't taken off," Christy says. "We're sure it will."

The impetus will be Web-enabled applications that have a high business value, he says. "It won't be to accelerate access to ESPN scores."

Other industry watchers aren't so sure. Enterprise customers will only deploy Web caching in a big way when it's free, says Ted Julian, who recently left his position as research analyst with Forrester Research to join Battery Ventures, a venture capital firm in Wellesley, Mass., as a start-up advisor.

He's only half-kidding. Julian expects equipment providers will build caching into network equipment such as edge routers, making the capability just another feature or, at worst, a low-cost upgrade.

But overall, Julian says the enterprise market will be small potatoes for Web caching companies. "That's not to say there won't be an increase in volume for the enterprise, but it'll be a rounding error; it'll be anemic," he says.

One thing about Web caching is clear, however, and that is the type of cache that users want to deploy. The so-called cache appliance - a box that does nothing but Web caching - has won over more limited proxy servers and multipurpose servers. These appliances are available from companies such as CacheFlow, Cisco, IBM, InfoLibria, Inktomi and Network Appliance.

Prices are likely to come down as well. Whereas high-end cache appliances once sold for up to $100,000, companies including Novell have announced cache software that runs on a dedicated Compaq or Dell server at a price Christy estimates to be in the $40,000 range. "It's pretty hard to buy a $100,000 Dell server," he says.

Related links

Face-off: Is enterprise caching worth it?
Entera and InfoLibria debate the merits of caching technology.
Network World Fusion, 8/2/99.

Caching vendors to debut new software
Network World, 9/6/99.

Net Resources: Caching
Reviews, articles and other Web resources.

Cache cachet
Web caching appliances help conserve precious Internet bandwidth.
Network World, 5/10/99.

Waiting for Windows 2000

W2K talk
Network World columnists Tom Nolle, Dave Kearns and James Kobielus deflate W2K hype (4 min. in G2 RealAudio).

Windows 2000 is now scheduled to ship sometime late this year, but pragmatic IT shops won't be touching it until much later than that.

That's the sound advice megaconsulting firm Gartner Group has been giving regarding Windows 2000. While Windows 2000 will bring the long-awaited - and much needed - Active Directory, which promises to help you better manage big NT nets, it also promises to be fraught with bugs and myriad implementation "gotchas." Do you really want to do this to yourself so soon after you've (hopefully) put the Year 2000 issue to bed?

Admit it. You don't. And you shouldn't.

The reasons are many, says Thomas Bittman, vice president of research and advisory services at Gartner:

Windows 2000 is too expensive. The prices of computers may be going down, but the price of Windows software is heading up. For example, if you want to support four-way symmetric multiprocessing, you'll need Windows 2000 Advanced Server, which Bittman expects will cost roughly twice as much as the NT Standard Edition software that supports that capability.
More than half of all users will fail to achieve the expected costs savings of a Windows 2000 implementation through 2001 because of the high costs of hiring, training and retaining employees, and bringing in outside help.
Windows 2000 will not be reasonably stable and feature-complete until at least the first half of 2000. Additionally, it will be less reliable than Windows NT 4.0 until year-end 2001.

Those are just some of the highlights; Bittman and Gartner Group have many more reasons to be wary of Windows 2000. Bottom line, Bittman says: You should not deploy Windows 2000 at the desktop or as a midsize network operating system (NOS) until Microsoft delivers the "first major proven service pack." He expects that will be six to nine months after general availability of Windows 2000. For larger NOS deployments, Bittman recommends waiting a year.

Related links:

Recent Network World articles about Windows 2000
From the past six months, ranked by relevance.

What's up with Windows 2000?
Jim Allchin lets you in on the product's progress.
Computerworld, 8/3/99.

Ready or not, here comes Windows 2000
Network World, 6/28/99.

Windows 2000 information site
from Microsoft.

E-mail management mania

E-mail talk
Network World columnists Tom Nolle, Dave Kearns and James Kobielus deflate e-mail hype (3.5 min. in G2 RealAudio).

With the proliferation of corporate Web sites has come a deluge of e-mail generated by those sites. And with that has come a rather huge problem: Who's going to read all these missives, never mind respond to them?

The answer has arrived in the form of e-mail management tools from the likes of Aditi, Brightware, eGain Communications, G2X Software, Kana Communications, Mustang Software and Silknet Software, to name a few.

To varying degrees, these software tools can read incoming e-mail, route mail to qualified customer service representatives based on subject matter, and even compose and send meaningful responses to some messages.

What's truly amazing about this technology is that it really works. Using Brightware 3, Egghead.com is able to handle five times more e-mail with about half the staff it used to have, says Norman Hullinger, vice president of sales and operations for the online software retailer. With a staff that peaks at 15 people, and typically includes only 12 or 13, Egghead.com is able to field 900 to 1,100 messages daily.

Brightware answers 30% to 35% of those messages without human intervention, Hullinger says. Typically these are simple requests, such as questions about shipping policies to various states or countries. For more complicated queries, the software either composes a response that is then checked out by a human or it routes the incoming mail to the most appropriate customer service representative, depending on the content.

"This type of product is a must-have for any one with an active Web site," says David Cooperstein, research director for consumer electronic commerce at Forrester.

Ultimately, you can expect e-mail management products to get wrapped up in larger customer relationship management (CRM) systems. CRM tools track all interaction between a customer and a company, be it voice, phone, fax, snail mail or e-mail.

The CRM market is certainly much larger than e-mail management alone, although both have decent growth rates. Forrester expects customers to spend about $71 million on e-mail management products this year but $658 million in 2002. By contrast, last year users spent $1 billion on CRM products and Cooperstein expects spending to hit $3 billion by the year 2003.

Web site e-mail products maturing
Network World, 11/2/98.

PKI's time will come

PKI
Network World columnists Tom Nolle, Dave Kearns and James Kobielus deflate e-mail hype (8.5 min. in G2 RealAudio).

In a word, public key infrastructure technology is confusing. There is much to know and lots to go wrong if you don't adequately determine all the angles. To date, that has scared away all but the most determined, or perhaps most desperate, users.

That will change, however, because it must. When you're talking about electronic transactions worth tens of thousands of dollars or more, you've got to have assurances that the people on each end of the transaction are who they say they are. And you've got to be able to prove, for example, that a $100,000 order for pin cushions placed on Sept. 27 was indeed placed on that day should either side try to renege. And it's nice to know that nobody has tampered with the order in transit, too.

PKI gives you all that, but it's not easy getting there. First, you've got to have a certificate authority, which doles out the private and public keys to various users. Messages are encrypted with the public portion of a user's key and decrypted with the private portion. Your public key is generally accessible to anyone who might need to send you an encrypted message, but the private key you pretty much guard with your life.

The public key is also known as a digital certificate. It carries information that positively identifies you to the party on the other end - the certificate authority makes sure of that.

All these components are available from vendors such as CyberTrust, a GTE company; Entrust Technologies; and VeriSign.

Among the sticky PKI issues in various stages of resolution are digital certificate management, integration of PKI components with back-end applications involved with transaction processing systems, standardization of keys and private key storage.

Key management is an especially challenging one. Certificate authorities generally keep revocation lists of bad digital certificates, but checking against those only assures you that the certificate hasn't been revoked yet; it doesn't give you any assurance that it won't be revoked tomorrow and that the party in question will be able to pay for whatever is being ordered. That leaves the certificate authority carrying a lot of risk.

But companies are springing up to address such issues. One is Identrus, a PKI service provider in New York. The company has enlisted a group of banks to help it build a "global trust organization" that will manage electronic transactions and assume the risk involved.

The launching of Identrus and other such companies, along with continued work on standardization and open programming interfaces to ease integration, will likely make PKI more accessible to the masses in the not-so-distant future.

Related links

Are you usefully certifiable?
Bradner on PKI. Network World, 8/16/99.

The PKI Page
Links to PKI info.

Desmond, formerly Network World features editor, is now vice president of King Content, a custom publishing company focusing on information technology. He can be reached at paul_desmond@king-content.com.

Back to the Buzz home page

To top