Roadwork ahead

everal organizations are feverishly working to revamp the Internet into a communications speedway capable of supporting next-generation, bandwidth-hungry multimedia applications.

If all goes as planned, a fortified Internet could emerge in approximately three years to five years. But much hinges on whether currently disparate efforts such as academia's Internet2 (I2) and the U.S. government's Next-Generation Internet (NGI) program can successfully coordinate their activities.

The new era of Internet applications, which will quickly devour bandwidth and require a consistent set of end-to-end, network support services, have the potential to revolutionize teaching and research practices. They also will transform traditional business-commerce models, medicine, training, and everyday communications among business associates, family, and friends.

Among the applications on the horizon are networked virtual reality for training and collaboration, downloadable audio and video libraries, and remote scientific modeling.

GETTING GOING. To support these next-generation applications, the new Internet will require a reinforced network infrastructure with much higher speeds and a consistent set of middleware services. Getting these services in place does not depend on the creation of new technologies so much as it requires the fleshing out and standardization of schemes already under development by commercial companies.

"For example, we are working toward consistent implementations of QOS [quality of service], security mechanisms, directory services, and video encoding networkwide," explains Ted Hanss, director of applications for I2, in Ann Arbor, Mich. I2 is a project involving 120 U.S. universities, plus government and industry partners, to accelerate the next stage of Internet development in academia.

Hanss elaborates: "In terms of security, we're not trying to develop new encryption techniques. But we do need a system that does not require users to each have more than 100 log-ins to access servers on each I2 campus."

Similarly, Hanss adds, "We don't need a new form of video encoding, but we do need a way to make what we already have scale. Today, video plug-ins for different Web browsers are encoded in different ways, and most client browsers are optimized for dial-up connections. They don't really exploit the full-motion, full-screen video that is possible across a gigabit-speed network."

THE ROAD TO ABILENE. The most notable efforts for building a next-generation Internet involve I2 and the federally funded NGI project. I2 is relying on a minimum of one and one half-million-dollar contributions by university participants, as well as some federal funding and the backing of industry partners, such as AT&T, Advanced Network & Services, Bay Networks, Cabletron Systems, Cisco Systems, Fore Systems, IBM, Lucent Technologies, Nortel, 3Com, and others.

There also are state-government and local-government NGI projects sprouting up, as well as efforts in other countries.

The hope is that with industry involved from the start, the commercial world will benefit more quickly from government and academia's research and progress in advanced networking than it did with the original Internet.

The I2 test network, which will be shared -- to a degree -- by the NGI, was announced in April by U.S. Vice President Al Gore. Dubbed Abilene, this network borrows capacity from network service provider Qwest Communications' fiber-optic network.

Abilene consists of regional traffic aggregation centers called gigaPOPs (gigabit Points of Presence) that provide ultra-high-speed connectivity among universities and other organizations. These gigaPOPs offer connectivity at about 100 times the basic backbone speed of today's Internet. Through them, universities will connect to NGI networks and other advanced federal networks. Full deployment is expected by the end of 1999.

SEPARATE PATHS. One area of potentially problematic divergence, though, is that I2 has standardized on ATM as its physical infrastructure, while NGI is leaning toward Synchronous Optical Network and Gigabit Ethernet. Disparate physical-layer platforms could pose problems for delivering consistent end-to-end QOS, industry experts say.

"If user traffic must traverse `islands' of a next-generation Internet, there will be no guarantee of QOS across all the islands," points out Atul Kapoor, managing director of the Tolly Group, in Manasquan, N.J. "This defeats the purpose of QOS."

The Tolly Group, known for its independent, network-product benchmark testing recently formed a consortium called the NGI Forum after large end-users such as Goldman-Sachs, Virginia Tech, and Electronic Data Systems appointed the company as a neutral third party to coordinate disparate NGI efforts.

Finding a standard for QOS, a scheme that will ensure consistent network capacity and prioritization for delay-sensitive traffic that requires it, has long been a stumbling block for IP networks. This is because IP is a "connectionless" protocol that does not set up a predefined path for data to take across a network. Nailing down a QOS standard for IP networks is on the I2 and NGI agendas.

"Another problem is that there are no uniform agreements on what traffic should take priority on a public network," observes Steve Wolff, head of the Division of Advanced Internet Initiatives at Cisco Systems, in Washington, and a member of I2. "You may know how to prioritize traffic through your own private network. But who decides who winds up first in the queue in a network used by many organizations?"

GUARANTEED DELIVERY? Once traffic has been prioritized, the network must be capable of delivering the specified bandwidth and QOS requirements.

"ATM is the only network today that inherently specifies a way to deliver QOS," asserts Phil Vermeer, an NGI Forum representative from the end-user community.

Vermeer, a member of the U.S. National Guard Director's Future Planning Group, in Arlington, Va., is interested in the use of next-generation networks to support 3-D simulations.

ATM supports QOS through its five classes of service and its inherently circuit-oriented architecture. On the other hand, Gigabit Ethernet has no QOS built into it. To deliver QOS over Gigabit Ethernet, additional technologies are required, such as Resource Reservation Protocol (RSVP), a technology for reserving end-to-end bandwidth, which was developed by Cisco.

"Gigabit Ethernet bases its access mechanism on collision detection and avoidance, which, by definition, is random," Vermeer continues. "Schemes that run on top of it, such as [RSVP], are proprietary."

In addition to its proprietary nature, RSVP has met some technical criticism. It allows users to reserve bandwidth only so long as the bandwidth is actually available end to end, from source to destination across the network, at the moment an application initiates a call. If the required bandwidth is not available along the entire route, the concept of a service-level guarantee goes out the window.

Also, in a routed network like the Internet, traffic dynamically changes routes to its destination. Without a known path, it is difficult to reserve resources. Standards work is afoot, though, on a technology called Multiprotocol Label Switching (MPLS) that could help remedy this QOS situation. MPLS combines routing and switching in WANs in a way that defines the traffic path. RSVP will likely perform path setup functions for MPLS.

WHO'S IN THE DRIVER'S SEAT? Some industry observers look askance at the whole I2 and NGI movements, largely because of the federal government's involvement. Unlike the original Internet, which was in and of itself a government-supported entity that wound up serving the commercial sector, today's Internet already has a large commercial foundation. Some observers cringe at government involvement in what has become a commercial enterprise.

"I'm skeptical of Internet2 and NGI," states Boston-based InfoWorld columnist Bob Metcalfe, who invented Ethernet and founded 3Com. "The universities seem to be saying, `The Internet is broken and needs to be fixed. Let's have the government buy us a new one.' I think they should buy their own."

Tom Nolle, president of CIMI, a network consulting and research company in Voorhees, N.J., adds: "Publicly funded carriers are not going to tolerate the government building their backbone to compete in a public arena, nor should they. Anyone who thinks otherwise is smoking dope."

VIRTUAL VOYAGES. Three-dimensional training and simulation seem to be among the most attractive early applications.

"We can train soldiers on how to change a brake-drum assembly on a Humvee," Vermeer says. "Traditionally we take the soldier to a shop several times with an instructor and have him practice with expensive tools on the real thing. By letting a solider practice over and over again in a simulation environment, we save expensive shop and instructor time and costly mistakes."

Wolff envisions the eventual use of an interface called The Cave, a virtual-reality application much like the holodeck in the popular Star Trek series. A 10-foot cubicle with a headset and 3-D projections on the walls, The Cave could virtually transport users to a trade show where they could attend a lecture, have real-time conversations, and experience the event just as though they were there.

However, The Cave would easily chew up a T3 circuit (45Mbps) on its own, Wolff says.

"And at $1 million a pop today, it will be a long time before everyone has one in the study," Wolff adds.

Joanie Wexler is an independent networking editor in Campbell, Calif. She can be reached via e-mail at joanie_wexler@mindspring.com.