Advertisement
Articles
Advertisement

Grey matter

Sun, 04/30/2000 - 8:00pm
Jeff Baumgartner, Senior Editor

Oh, if I only had a brain.

That jovial refrain, as we all remember, comes from the Scarecrow in the film adaptation of L. Frank Baum's wonderful children's yarn, "The Wonderful Wizard of Oz," as the strawed-one, Dorothy, Toto and their gang of needy cohorts battled flying monkeys and other evil ilk as they rambled down the Yellow Brick Road in search of the Emerald City and, ultimately, the Great Wizard himself.

i_may2a.gif
ICTV's headend-centric topology. Courtesy of ICTV

Of course, when it comes to the cable industry, which is traveling its own road of evolution armed with the lucrative promises of two-way, broadband networks and a multitude of interactive services like video-on-demand, high-speed data and time-shifting personal video recorder (PVR) capabilities, the question today isn't whether those networks require a smart noggin—it's where to put the darn thing.

Debate on that very subject is growing, some with more extreme attitudes about the issue than others. For instance, companies like WorldGate and ICTV support the notion that a healthy portion of that intelligence should sit inside the headend coupled by a thin client in the home, while others believe most of the processing power should be contained within a thicker set-top box equipped with a heady, ultra-fast The broadband network intelligence debate continues to brew Oh, if I only had a brain.

That jovial refrain, as we all remember, comes from the Scarecrow in the film adaptation of L. Frank Baum's wonderful children's yarn, "The Wonderful Wizard of Oz," as the strawed-one, Dorothy, Toto and their gang of needy cohorts battled flying monkeys and other evil ilk as they rambled down the Yellow Brick Road in search of the Emerald City and, ultimately, the Great Wizard himself.

Of course, when it comes to the cable industry, which is traveling its own road of evolution armed with the lucrative promises of two-way, broadband networks and a multitude of interactive services like video-on-demand, high-speed data and time-shifting personal video recorder (PVR) capabilities, the question today isn't whether those networks require a smart noggin—it's where to put the darn thing.

Debate on that very subject is growing, some with more extreme attitudes about the issue than others. For instance, companies like WorldGate and ICTV support the notion that a healthy portion of that intelligence should sit inside the headend coupled by a thin client in the home, while others believe most of the processing power should be contained within a thicker set-top box equipped with a heady, ultra-fast processor. Still others call for a generous mix of both, or the use of a brainy residential gateway that's mounted to the side of a house.

As the thick- or thin-client deliberations continue, there's also plenty of discussion on the issue from chip-set suppliers, as well. One company that is making some noise is Equator Technologies, which, in another shot across the bow of Broadcom and its market-leading hardwired chip-set approach, recently introduced the MAP-CA, a third-generation, software-based processor. And, don't look now, but even Big Blue itself, IBM, claims to have built its own "It'll-do-everything-but-mow-the-yard" set-top processor.

Client intelligence: Is thin in?

One of the livelier debates regarding network intelligence is occurring in the client-server versus in-box solution side of the business. The apparent leader of that debate continues to be WorldGate Chairman and Chief Executive Officer Hal Krisbergh, the very vocal advocate of a headend-centric, "ultra-thin" client architecture for Internet-over-TV services.

"Thin client is our story," he says. "That's been our story consistently. We've had some people try to copy us like Peach (Networks Ltd.), Source (Media) and ICTV, but we are alone, way out in front, and making it happen."

Indeed, Krisbergh, who had to weather his share of heat on the issue in the early going, is beginning to make some serious hay in terms of deployments both in the United States and abroad. In fact, WorldGate currently has 40 deployment deals under its belt, and ended 1999 with 15,000 subscribers, up a lofty 32 percent versus the previous quarter.

In addition to making his presence felt at industry gatherings and inside boardrooms, Krisbergh is also stating his case on the national stage. Pushing his own "Digital Divide" remedy, Krisbergh grabbed the attention of ABC, NBC and other large media outlets in March when WorldGate partnered with Charter Communications to offer free Internet access via television to folks in LaGrange, Ga. by marrying Motorola's line of DCT-2000 set-tops with WorldGate's EVERY TV Service. Following the announcement, demand for the free service in LaGrange sent "the phones ringing off their hooks," he says.

Putting the Internet in every home is Krisbergh's latest initiative, claiming such a move would raise enough incremental revenue for it to be worth cable's while. To spur such missionary work, Krisbergh plans to harness SURFview, a $99 Internet-TV set-top introduced last year with General Instrument (now part of Motorola's Broadband Communications Sector).

"Internet on the TV related to TV programming," he argues, "can generate somewhere between $10 to $12 in incremental revenue for the cable operator. If people really accept that number, and the cable operator has the ability to put a low-cost box like SURFview in the home that only costs $1 a month for the cable operator to amortize, you would put a box in every home. Putting a box in every home is by far the single biggest opportunity that the industry has . . . period," he says.

While the idea of putting a box in every home remains unsettled among most MSOs, the same is true for cable operators when it comes to employing the majority of a network's intelligence via the headend-server or in-box model, he says. "The headend-server approach has been a cable approach all along," he says. "S-A (Scientific-Atlanta) and GI have been doing that for years. The boxes they've used are very capable, but the intelligence was leveraged on the cable infrastructure. That way, you could use a headend where the cost of memory and CPU is very cheap to serve a lot of boxes, rather than trying to put a box in every home with a lot of CPU and memory in it."

Such a concept, Krisbergh says, is classic networking. "Cable uses it, telephone uses it—it's what makes sense," he deducts. "The PC didn't use it because they were too focused on computing on a desk. Once the desktop computer was there, it was logical for them to leverage that resource to do communication."

One player that is trying to come up WorldGate's thin-client tailpipe is ICTV, which hit the reset button last November and launched a new, headend-based, digital platform based on MPEG-2 to link cable customers to broadband content via all current and planned set-tops without the need for plug-ins or middleware.

Admittedly late to the game when the platform was introduced, the fruits of ICTV's labors are finally starting to bloom. St. Joseph Cablevision in Missouri recently agreed to deploy ICTV to its 10,000-home system, and two former Bresnan systems (now owned by Charter) that pass a combined 60,000 homes in Bay City and Midland, Mich. rolled out ICTV in March. What's more, the company, which has been targeting top 10 MSOs in the U.S., expects to announce more platform deployments during (ta-dah!) this month's National Cable Show in The Big Easy. By managing the service at the headend, ICTV believes it can help cable operators address daunting lifecycle issues for digital, interactive services.

"If ICTV and an advanced digital set-top come out of the gate at parity and both deploy Pentium III 600s," says Michael Collette, ICTV's senior vice president of marketing, "two or three years from now, that thing will be a dog, and we're both going to have to upgrade. The difference is that we'll upgrade at the headend and just pop it in."

That difference also can ease a cable operator's bottom line, he adds, because the cost to swap out a box is in the range of $300, plus a truck roll. "Even if that upgrade costs $1,000, if we amortize it over 100 customers, it's $10 per customer rather than $300," argues Collette.

While intelligent headend solutions like ICTV's could save on cable operator upgrade costs, the ability to scale such a service continues to be a key question. ICTV has that covered, too, says Collette, because ICTV only needs to provision for peak usage. That way, equipment doesn't languish on the racks, unused, he adds.

"If someone churns, we just turn them off," he says, using a model of 10,000 customers. "We don't lose equipment of any kind. With 10 percent of those homes using the service, we have to service 600 simultaneous sessions, which would suggest that we need 600 processors. But some of those are only light duty things like e-mail and browsing, allowing us to multitask the sessions. With light duty, we think we can get about 20 simultaneous sessions on a single digital control module, which is basically the processor."

While WorldGate and ICTV continue to make inroads with their respective solutions, a smattering of recent moves by Microsoft Corp. and Liberate Technologies suggest that the thin client argument holds more water than originally believed.

Microsoft struck a $75 million deal in March to take over Israel-based Peach Networks Ltd., giving the software giant a way to expand its higher-end Microsoft TV platform with one that can be deployed to the millions of digital set-tops that sit in homes today.

Meanwhile, in January, Source Media and Insight Communications sold their jointly owned VirtualModem technology to Liberate in exchange for equity positions in Liberate. The deal, the companies said at the time of the announcement, would extend the Liberate software platform to a broader range of set-top boxes, including DCT-1200s and DCT-2000s.

"There are all sorts of networks out there. Everyone's different," says Dave Limp, senior vice president of corporate development at Liberate.

Limp adds that the current status of deployed broadband networks will continue to define the thick or thin question of the future, and that ultimately every company will need to have products that can meet a mix of scenarios. "It's important that we have solutions available that can be handled by any infrastructure."

Krisbergh, however, maintains that thin client is much more than an interim solution, and such a view is a major source of frustration with him. "It's very clear that the thin client architecture is prevailing," says Krisbergh. "However, no one has abandoned the thick client solution totally yet. Thin is prevailing, but it's not yet settled."

Mixing it up: The applications factor

While WorldGate and ICTV represent the extreme on the side of thin-client solutions, others believe generous amounts of a network's intelligence should inhabit both sides of the network, depending on the application in question.

"(WorldGate's solution) works for a lot of customers, primarily those today who don't have a PC in the home," says Tom Jokerst, senior vice president of advanced technology at Charter, a WorldGate customer.

"I think in the short-term we can take advantage of the network capacity and utilize it to its fullest extent," he says. "Long-term, we're going to want set-top boxes that have storage so that we can continue to accommodate more features and functionality to the end users."

When TiVo-like services become big business, he adds, cable operators will need to re-scale their networks or put that functionality on the client side.

Bill Wall, S-A's technical director of subscriber networks, agrees that determining where the brains of a broadband network should sit is highly application dependent.

"We certainly believe you need a relatively powerful home client device," he says. "When it comes to scaling of general applications like Web browsing and navigation functions, it makes sense to have that intelligence distributed out to the home," he says. "If you try to do everyone's Web browsing at the server side, then I think you start to run into bottlenecks."

While Wall believes a thicker client would better aid applications like Web browsing, others like video-on-demand, for instance, work extremely well in a thin-client environment, he says

"The amount of processing needed in the set-top is very minimal for that sort of event. It makes a lot of sense to do it that way," he says. "However, for a customer to get the full spectrum of applications, he's going to require a fairly intelligent set-top in any event."

One of those applications is the popular MP3 music format, he says. "You could make the argument that you could decode MP3 in the headend, and send that as an uncompressed, digital audio stream down to the set-top, but that's going to eat up an enormous amount of bandwidth," explains Wall.

AT&T Broadband, which has yet to deploy wares from WorldGate or ICTV, also agrees that a blend of network intelligence is in order as new applications and services evolve. "There's always going to be a big need for us to support thin-client just with the number of DCT-2000s that we'll continue to support in the field," says Jim Wood, vice president of advanced technology at AT&T Broadband. "As we get a mix of boxes out there, we'll have some flexibility to either rewrite our services such as video-on-demand or e-mail to ride on a variety of different platforms. The vendors are going to employ a lot of innovation, but we want to provide services that will be able to take advantage of whatever hardware happens to be out there."

While the intelligence debate broils among premise and headend equipment, still many others see a future involving extremely smart residential gateways as home networking enters the fold.

"With home networking, we believe IDDs (Interactive Display Devices) will be used initially for controlling home automation, electronic programming guides and other things," says Andy Trott, Pace Micro Technologies' chief technical officer. "It's possible for it to be very intelligent and for it to have a high-power CPU in there at a very reasonable cost. It's just a content decoder for video. In a very limited form, I think we'll see something like that by the end of this year, but not something that will be carrying video."

Smart (and very smart) processors

As the network intelligence debate rages among service providers and box vendors, the chip-set manufacturers have their own intelligence battle brewing. The topic at hand tends to be whether primarily hardwired or fully-programmable chips provide the best processing solution for both existing and budding applications and services like PVRs, high-definition television and video conferencing.

Taking the software-based extreme is Equator Technologies and Hitachi Ltd., which recently introduced the MAP-CA, a third-generation media accelerated processor for consumer appliances that approaches upgrades via software downloads rather than hardware replacements.

"This chip is going to drive a stake in the heart of the hardwired devices and just obsolete them all at once," exclaimed John Setel O'Donnell, Equator's co-founder and president, when the MAP-CA was unveiled.

Such a declaration of hardwired obsolescence roughly parallels Hayes' downfall in the narrowband modem space during the early 1990s, O'Donnell says. "They were doing a great job of managing slow improvements in modem operating speed, but in 1991 Rockwell and US Robotics began shipping software-based modems where the modulation and demodulation functions were done on a DSP (Digital Signal Processor) in software." Spurred by such innovation, O'Donnell adds, the new kids on the block completely displaced the hardwired incumbents within a few short years.

"(In the set-top and television arenas), we're moving away from a platform where most of the silicon budget is spent on ASICs (Application Specific Integrated Circuits), to one where the silicon buys you a microprocessor and all of the TV functions delivered by software."

Perhaps second only to cold fusion, the software-based processor approach is "the Holy Grail, the dream of dreams, the future of futures, and the dream of mankind since the dawn of time," counters Tim Lindenfelser, vice president and general manager of Broadcom's Communications Business Unit.

"We looked at Equator two or three years ago," he says. "We asked them to show us what the MIPs (millions of instructions per second) would be, what the die size would be, and what the power would be. From there, we said we'd compare that to our hardware solution and see whether it reaches the cost trade off. It never did."

Programmable processors, adds Lindenfelser, have to be so speedy that by the time they are finally fast enough to do a hardware-based function, the hardware vendors have already moved on to the next function or next generation. "For instance, take MPEG-2," he adds. "That's a fixed function, so there's no longer a reason to make it programmable because it's a standard. You can now do MPEG-2 in software on a 600 MHz Pentium on a PC today, but that's a ridiculous number. Today, an MPEG-2 core is a small, little hardware design."

Lindenfelser admits, though, that the programmable processor concept resonates well with manufacturers and service providers. However, he points to Silicon Valley start-up MicroUnity Systems Engineering Inc. as a prime example of why the software-programmable notion has a daunting set of challenges in front of it.

Formed in the early 1990s and backed by TCI (now AT&T Broadband), Time Warner, Cox, Comcast and a number of equipment vendors, MicroUnity's early activities centered on the work being done at MCNS (Multimedia Cable Network Systems), which helped to pave the way for the interoperable DOCSIS specifications that followed.

"TCI, Time Warner and a number of other cable operators and equipment makers invested $200 million in MicroUnity with the same dream of creating the super-dooper-wooper, software-does-everything processor, and it was just an unbelievable failure," says Lindenfelser.

MicroUnity's problems, he explains, came about because programmable chips require enormous amounts of gates and circuitry. "When you make it hardwired," he says, "you can intuitively believe that it would be smaller and have less circuits. If you make it with software, you've got to have a lot of redundancies to make it programmable."

Though MicroUnity's early efforts might have been a little before their time, counters MicroUnity Chairman John Moussouris, the work his company completed in the 1990s has helped to usher in the dawn of economically- and technologically-viable software-based chipsets.

"When we began working on this problem in 1993, it was of great interest to the service providers, the MSOs and others," he explains. Moussouris adds that it was also true at the time that many more transistors were required to solve the problem using the techniques of the day. "Back then, the manufacturing base wasn't ready. We did a lot of research to try to get a jump on Moore's Law."

That lithography (the method of printing smaller and smaller chips) research started paying dividends when Micro-Unity licensed its technology to PC chip vendor AMD and TSMC (Taiwanese Semiconductor Manufacturing Corp.), one of the largest independent manufacturers of silicon in the world. What's more, MicroUnity is still working with a number of broadband equipment vendors, though the names of those companies remain confidential. Nevertheless, those business transactions have helped Sunnyvale, Calif.-based MicroUnity sweep away all of its debt, says Moussouris.

"People argue that (programmable) chips will be bigger and will need higher power," he adds. "That may be true at one stage in history, but the gap narrows. I've been in this industry 25 to 30 years, and it has always been the case that functions that are initially supplied by digital hardwired get supplanted at some point in time by digital programmable."

After surviving the usual strife associated with trying something new and untested in the early 1990s, Moussouris remains a true believer in the potential of programmable processors, and the positive economic impact they could make on a service provider's bottom line.

"Upgrading the ASICS is an engine in (Broadcom's) business model, and they're doing very well at it," he continues. "Upgrading of ASICS is a wonderful cash cow for them. They will probably try to preserve that for as long as they can."

While Broadcom takes on a hardware approach with its processors, the company doesn't advocate an extreme on either end of the hardware/software spectrum.

One example is the Propane software Broadcom acquired this year from Digital Furnace, which calls for software elements to rest in the premise and at the headend. There is also a level of programmability required for PixelSquirt, a 3D graphics engine that will be added to Broadcom's chipsets. Broadcom struck a deal earlier this year to acquire PixelSquirt's developer, Stellar Technologies.

"Our approach is more one of moderation rather than extreme," says Lindenfelser. "We'll hardware what's standardized, but make other parts programmable for things that aren't completely standardized."

Meanwhile, Equator has taken its approach to task, and has signed on several backers of its SofTV initiative, including Le Groupe Videotron, which has built a residential gateway in Quebec, Canada; RealNetworks, Ward Laboratories, Wind River Systems and Zydacron Inc. Additionally, the company has seen the price of its processors dip from the hearty sum of $200 to $500 per unit for its first-generation processors, to a mere $40 apiece in quantities of 100,000.

Programmable players are growing

Of course, Equator isn't the only one waving its arms about programmable processors.

IBM just introduced a family of chips for set-top boxes designed to drive new capabilities for EPGs, Web browsing and interactive applications such as home banking, e-commerce and information retrieval. IBM's PowerPC 405 "system-on-a-chip" will become available this month, and a lower-end PowerPC 401 is slated to come off the lot in the second half of 2000.

In addition, ishoni Networks introduced its $30 to $60 "gateway-on-a-chip," which uses a software suite that combines voice, networking and security processors. OEMs that use ishoni's chips will have the ability to integrate their own intellectual property, the company says. ishoni started shipping samples this month, but declined to disclose its customers/partners.

Time, as usual, will tell

The best technology doesn't always win in the marketplace—VHS clobbered Beta, Windows continues to beat Mac OS, and so on and so forth. Today, the network intelligence debate is still far from being decided—the industry still doesn't know what will work the best on a uniform basis.

Like most consumer electronics, the technology behind it doesn't always drive product sales and make it reach critical mass all on its own. In addition to good ol' know-how, the factors that will determine the winner (in any industry) will also include a steady mix of word-of-mouth, PR, sales and marketing, and maybe even a little arm wrangling.

The real race is on in the intelligence debate. Who wins is still anybody's guess. But it will be fun, nonetheless, to watch the players and see which ones make it to the end of the Yellow Brick Road first.

E-mail: jbaumgartner@cahners.com

Advertisement

Share This Story

X
You may login with either your assigned username or your e-mail address.
The password field is case sensitive.
Loading