Networking's Future: Industry Innovators Speak Out

As the networking market emerges from the nuclear winter of the recent telecom crash, the question arises: Is innovation still a part of networking? And if so, where will it come from?

For the answer, we turned to experts in the field, who are largely optimistic about the future of innovation. However, that enthusiasm is tempered by predictions of tougher capital and spending markets, as well as harder work necessary to find niches where innovation could lead to product and company success.

So while the days of companies being founded on nothing more than a good idea might be past, there's still room at all points of the network infrastructure--from the underlying chips to the software, protocols, hardware, and applications--for innovation to deliver new services and perhaps new players to the networking arena.

Networking innovation, our experts agree, is needed to deliver the advanced services of the digitized future, from more reliable and more easily deployed broadband networks to feature-filled services for consumers and businesses, such as streaming video and VoIP.

All along the network's path, new technologies will be needed, starting at the chip and protocol level and proceeding up to the servers and routers that networks run on. Last but not least, there must also be innovative thinking on funding and the creation of new businesses, a prerequisite to surviving in a world where investment capital is scarce and increasingly scrutinized before it's parceled out.Our distinguished panel of experts include Internet co-founder Bob Kahn, who also co-authored the TCP/IP stack; serial entrepreneur Judy Estrin, founder of a number of networking concerns; Broadcom co-founder Henry Samueli; VoIP impresario Jeff Pulver; and Avici Systems co-founder and product architect Larry Dennison. In the following series of interviews, these networking sages predict where innovation is likely to come from in the near and long-term future, and what the challenges are for those innovations to bear real fruit in the networking marketplace.

Paul Kapustka is editor of CMP's Advanced IP Pipeline, a news and analysis Web site for the telecom provider and ISP communities. Kapustka has been covering networking since 1991. Send comments to [email protected].Bob Kahn

New protocols are vital in enabling network innovation, but they take time to develop and gain acceptance.

Innovator: Bob Kahn

Current Title: President, Corporation for National Research InitiativesClaim to Fame: Internet co-architect, TCP/IP co-author

Innovative Interest: Digital Object Architecture, a standard method to define and manage information on the Internet

You'd think that people would rush to follow the wisdom of one of the Internet's inventors, especially when he's proposing a standard protocol that could significantly improve the way digital objects are identified and administered on the Web.

But from personal experience, as Bob Kahn can tell you, there's no clear path to success for those looking to create innovative changes for the Internet. Even the best technologies might languish for years before they become widely accepted.

One of the Internet's founding architects, Kahn knows too well how long it takes for something to become an "overnight" success. A good example is TCP/IP, which Kahn co-authored with Vint Cerf in 1973. Only after years of persistent pushing did the interoperable open standard prevail over existing proprietary solutions such as IBM's SNA to become the lingua franca of the Internet.But just as he did in the Internet's early days, Kahn continues to push quietly for more innovation, always with an eye toward openness and interoperability. As he explains his points, Kahn can come across as a slightly avuncular professor, quick to help a befuddled reporter by sketching diagrams on his notepad. Yet his academic tone is tempered with a quiet, "knows-more-than-he-lets-on" presence, born perhaps of Kahn's 13-year stay at the U.S. Defense Advanced Research Projects Agency (DARPA).

Now president of the Corporation for National Research Initiatives (CNRI)--a nonprofit organization he founded in 1986 to "foster research and development for the National Information Infrastructure"--Kahn's current pet project is the Digital Object Architecture, an open architecture approach to rethinking the Internet that stresses managing information instead of just moving bits.

A key aspect of this architecture is the notion of unique persistent identifiers, which he refers to as "handles" and which are widely used in the publishing industry as a standard method for identifying and describing Internet-based content. The handle addressing system is somewhat similar to DNS addresses, with prefixes and suffixes that allow handles to be mapped to the appropriate administrative databases. Handle systems are also designed to interoperate with the current URL system.

"People need to know what you can and cannot do with information on the Net," says Kahn, explaining part of the need for unique persistent identifiers that identify individual data structures and other network resources independent of their location. A standards-based infrastructure based on the Digital Object Architecture could support many add-on applications, he says, such as one to help protect intellectual property, or to outline the terms and conditions for access to such information.

"This approach doesn't mandate enforcement," says Kahn, illustrating this point by comparing it to highway speed limits. "You know you're supposed to go 55, but there's no magnet in the sky that grabs your car and takes it to a smelter if you speed." Copyright enforcement applications can be built by others, he notes. "But first we have to come up with an underlying architecture that supports it."CNRI, which developed the handle technology that most Digital Object Architecture projects use, has been promoting the idea of handles for at least a decade, says Kahn.

Such an architectural concept seems like it might be readily accepted by an Internet community tormented by copyright issues, but Kahn says the Internet's own anarchic roots can also work against change, just as the phone companies resisted building the Internet more than three decades ago.

"Sometimes [innovation] works if you become part of the process," says Kahn. "Or you just work something out by yourself. Create a parallel universe, define some minimal levels of connectivity, and let that parallel universe happen."

And then get ready to wait and wait and persist--until your innovation is finally ready to become an overnight success.Henry Samueli

As networks converge onto a common IP fabric, similarly converged chips will be key to powering those systems.Innovator: Henry Samueli

Current Title: CTO and Board Chairman, Broadcom

Claim to Fame: Founder of Broadcom and PairGain Technologies

Innovative Interest: Converged-functionality chips to power the convergence of networks to IP

If the dream of a converged IP network is to be realized, convergence onto multifunctional chips is going to be imperative. So says Henry Samueli, CTO and co-founder of Broadcom.One could excuse Samueli for being biased in his view. After all, Broadcom is a premier supplier of all kinds of communications-related integrated circuits. The company is so busy that Samueli can't spend more than a few minutes on the phone without being repeatedly interrupted--this, at 5p.m. on a Friday afternoon, when other high-tech companies are heading for their traditional beer breaks.

"There's really no end to the amount of innovation in front of us," says Samueli. Broadcom's own innovations this past year include a chip that blends a Gigabit Ethernet controller with smarts for TCP offloading, Remote Direct Memory Access (RDMA) clustering, iSCSI storage, and remote system management, giving server manufacturers a step up in building network-aware systems. Such products are just the beginning of what Samueli sees as a new wave of innovation in networking and communications chips, supporting a move by all types of digital communications to an IP fabric.

"I know everyone talks about it, but convergence is very real," says Samueli. "CIOs everywhere are trying to converge their networks, and there are very exciting innovations ahead in that space."

According to Samueli, one of the technologies driving convergence is VoIP, viewed by many as the key to streamlining corporate infrastructures that today typically support separate voice and data networks. Storage also presents an opportunity for convergence, he says. As an example, he points to the advancement of technologies such as iSCSI over Ethernet, which holds the promise of being cheaper and easier to use than current storage solutions such as Fibre Channel.

Another integrated circuit-based networking innovation is RDMA, which allows clusters of servers to act as a single computer without having to deal with TCP stack issues, thereby improving overall system performance."All these things are converging onto a common IP fabric," says Samueli. "We've really just started today. These are just the first steps."

Samueli admits that most of the future's finished innovative products are likely to come from established players such as Broadcom, which have the necessary resources to design complex chips with tens of millions of logic gates of embedded computation. However, he adds that innovation for discrete parts of such future chips could come from start-ups with a specific focus or vertical specialty.

"Clearly, there's a role for start-ups to develop intellectual property used specifically in one area of a massive chip," says Samueli. "We're always looking for start-ups with an interesting piece of IP."

To that end, Samueli isn't just tire-kicking--he's helping seed the process, with contributions such as his recent $30 million donation to UCLA's School of Engineering, where he's still on "permanent leave" as a professor.

Samueli says he hopes the donation will give the chance for future Broadcoms to get started, but in the meantime the current Broadcom plans to keep innovating and developing new ideas of its own."It's fun to be innovative and in an environment where we can have a major impact on the broadband industry," says Samueli. "Designing chips gets more difficult all the time, and there are challenges. But we keep hearing that there's a brick wall, and that Moore's Law has hit the end of the road. I think that's still a ways off."Judy Estrin

It's time to look beyond the last-mile problem and focus on solving problems in the "first inch" of networks.

Innovator: Judy Estrin

Current Title: Chairman, Precision I/O; Chairman, Packet Design

Claim to Fame: Co-founder of several companies, including Network Computing Devices, Precept Software, and Packet Design; former CTO of Cisco SystemsInnovative Interest: Solving bottlenecks in server I/O subsystems

Her career may be described as being that of a "serial networking entrepreneur," but Judy Estrin says it's getting harder to follow that path, especially as networking matures and investment in research declines. However, Estrin is quick to point out that the maturation of networks itself has created a need for innovation to help alleviate bottlenecks in such areas as server I/O processing and mobility.

"Innovation is happening, but you have to work much harder to make it happen," says Estrin. That includes conquering the forces working against the funding of innovative research, which she says can also hamper progress.

That Estrin is able to see both the technical and business sides of innovation shouldn't be a surprise. In addition to her long career as a networking entrepreneur--having founded or led a list of companies that include internetworking leader Bridge Communications, low-cost workstation supplier Network Computing Devices (NCD), IP video pioneer Precept Software, and her current home of multiple networking concerns, Packet Design--Estrin also sits on the boards of the Walt Disney Company and Federal Express, giving her a unique vantage point of big-company economics.

Estrin herself brought up the question of funding innovation, tacking it onto an interview that started with the technical side of networking's future."I have a concern about how innovation is going to happen in today's environment," says Estrin. "How will it be funded? How will [companies] be motivated to work on innovative things, instead of the next product cycle?"

Estrin faced such a dilemma herself several years ago when she left Cisco Systems after serving as the company's CTO, a post she attained after Cisco purchased Precept. Her solution was the formation in May 2000 of Packet Design, an incubator research house that has now spun out three separate companies: Packet Design, which sells network route analyzers; Vernier Networks, which builds gateways to help secure wireless networks; and Precision I/O, a start-up with technology designed to improve the networking performance of servers.

Estrin, currently chairman of Precision I/O and Packet Design, is also on the board of Vernier.

Precision I/O focuses on solving what Estrin calls the "first inch" problem--that concerning the movement of packets on and off processors. As the world becomes more IP-centric and demands more server interaction, it also needs new architectures, says Estrin.

"We don't have an efficient server I/O architecture today," she says. Precision I/O, like other competitors in the space, is attempting to solve the problem with technology designed to offload some of the network processing from the central CPU. As applications such as VoIP become more mainstream, solving such network latency problems will take on even greater importance, she says.Longer term, Estrin sees a need for networking innovation in the area of embedded sensors. The reason is simple: As computers become ubiquitous in almost every space imaginable--from cars to handheld music devices to household appliances and beyond--they're going to need to communicate.

"Technologies like RFID [Radio Frequency Identification] are just the beginning," says Estrin. "That type of networking will take a whole new architecture to support low-power devices with lots of communication. I think you'll see that come out of academia in about five years or so."

Estrin is confident about one thing: "There will always be innovators, futurists, and people who figure out how to make things happen." However, they may not be overnight successes, she cautions.

"There will be a need to focus on incremental innovation," she says. "Innovation will still happen--just at a slower pace." Jeff Pulver

For VoIP to become a mainstream part of communications, the industry must learn to focus more on feature-based innovation and less on price. Innovator: Jeff Pulver

Current Title: President and CEO, Pulver.com Enterprises

Claim to Fame: Founder of Vonage and the Voice on the Net (VON) trade shows; co-founder of Free World Dialup

Innovative Interest: VoIP over Wi-Fi and new VoIP software features

One of Jeff Pulver's current missions is to find someone who can market VoIP to the masses--or, as he puts it, "be a Steve Jobs for VoIP." The irony of this pursuit is that in the interim it's the hyperkinetic Pulver himself who's emerging as the de facto Johnny Appleseed of the voice-over-broadband revolution.And his central point of contention is that to thrive VoIP providers need to innovate on the services and features side, instead of focusing simply on lower pricing compared to traditional voice services. "Voice over IP is a zero-sum game now, with everybody matching low prices," says Pulver. "If the VoIP industry is only known as cheap local service, we all lose."

While there are those who would say that Pulver is no slouch at promoting himself as "Mr. VoIP," his enthusiasm for new technologies and innovation in the broadband-enabled telephony space is real. The same geeky high-school kid who once crossed a ham radio with a cordless phone to give himself an early version of mobile voice capability, Pulver now oversees a small empire of voice-related business concerns that include the Voice on the Net (VON) trade shows, the largest VoIP-centric event in the industry, as well as software and hardware enterprises all centered around IP-enabled communications.

Rather than just pitch VoIP as a cheaper way to make phone calls, Pulver sees opportunities for innovations that can open up entire new streams of revenue. On the enterprise customer side, he sees VoIP over Wi-Fi as a technology that could exact an enormous potential for cost savings and operational control, if you believe research that says most corporate cellular calls are between phones on the same campus.

"Wouldn't it be cool for enterprises to take advantage [of Wi-Fi VoIP] and take control of their own cellular bills?" Pulver asks.

On the innovation front, Pulver sees near-future opportunities for Wi-Fi mesh networks running VoIP, as well as technologies that support VoIP roaming between Wi-Fi and GSM networks.Other areas of possible VoIP innovation include technological solutions for compliance with the Communications Assistance for Law Enforcement Act (CALEA, the FBI wiretapping regulation) and support for e-911 dialing services, neither of which is inherent to most VoIP implementations today. In typical Pulver fashion, he's doing more than just talking about the ideas: At his company's offices in Melville, NY, he says he's supporting five start-ups in an incubator fashion.

While those companies are already mining the VoIP market, in the near future Pulver predicts it will be someone from "outside the current industry" who will truly lead the commercialization of VoIP to bring it to the masses. Jobs, he says, offered the perfect example of how to do this via Apple's iPod rollout, which popularized digital media compression in a way none of the previous MP3 proponents could.

"The iPod wasn't new technology," Pulver notes, and neither is VoIP. But what broadband voice needs, he says, "is a personality with vision and execution, who can take a geeky technology and consumerize it."

While he searches for that spokesperson, Pulver is either busy fielding calls from top-tier venture capitalists, or writing blog entries about his appearances before Congress and the FCC (or maybe just about how he loves flying on Jet Blue). So even as Pulver says he doesn't know who will lead VoIP to the promised land yet, it's a good bet that whoever that person is will definitely know him. Larry Dennison

The advanced IP services of the future--including video and gaming--will depend on the core router innovations made today.Innovator: Larry Dennison

Current Title: Vice President of Architecture, Avici Systems

Claim to Fame: Co-founder of Avici

Innovative Interest: Designing core system routers to support advanced services

Innovation is still happening at the network core, according to Avici's Larry Dennison. The key, he says, is tailoring that innovation so that system vendors' customers--and their customers--can reap the benefits of that innovation to improve their own businesses. This means improving the reliability and ease of use of IP-based systems for the network's core."It's important to recognize what phase of innovation you need to be in," says Dennison, co-founder of the core router vendor and currently its vice president of architecture. In the core routing market, the days of radical new architecture introductions are past, he says. What matters most now are innovations in scalability, reliability, and the ability to quickly provision new services.

Avici, under Dennison's watch, has become an innovator in this arena, specifically with its patented Reliable Alternate Paths for IP Destinations (RAPID) technology, which uses algorithms to speed up recovery from node failures by using alternate paths across the network. Similar innovation from other vendors will help IP systems offer telephone network-like reliability, a must to support the SLAs that large customers require, he says.

"It's a lot different now than 1999, when it was just Cisco vs. Juniper," says Dennison. He says to survive, router vendors must also innovate in the areas of fast provisioning, allowing service providers to quickly turn services on or off to a customer site.

"There are two things we are seeing in all our new RFPs," says Dennison. "One is a demand for scalability and the ability to rapidly add capacity [to a router]. The second is reliability. Any core router now needs to have the same reliability as a class-4 or class-5 switch."

According to Dennison, the main driving force behind core routing innovation these days are the services offered by telecom providers and ISPs. New applications such as network-based VPNs and the emergence of VoIP will drive the need for innovation that's less about flash and more about cash--as in saving money for service providers while giving end users a feature that can help them run their businesses more efficiently.When it comes to underlying network needs, "VoIP is really a different animal," says Dennison. And current attempts at solving the demand for IP-based stability, such as the Resource Reservation Protocol (RSVP), aren't cutting it.

"If you're running VoIP, you want to have an ultrareliable network," says Dennison. "That's definitely where some of this 'hidden innovation' is going on. There's a pent-up demand for reliability [from customers] that wasn't there before."

Since it's a maturing market, Dennison doesn't see a wide opportunity for start-ups in the core routing arena--certainly not like it was just several years ago.

"The table stakes have kind of risen," says Dennison, who sees a future with no more than three core router vendors in the market. "Just having the basic protocol stack is something that's difficult to replicate," he says. "A single idea no longer makes it."

Of course, start-ups can also address some of the new demands on the core network, a point Dennison admits is part of his job at Avici."There is a whole new range of applications emerging, like online gaming, that is really very different," says Dennison. "Such new intense demands on the core of the Internet create opportunities for innovation--that's why we're always looking to see where the next stress point comes from."