FWIW: Internet2. I completely missed this when I asked for an overview earliers. This is from www.internettelephony.com
A Framework for Growth
By hosting next generation Internet projects, MCI WorldCom and Qwest hope to get a head start on technology that could make 3-D and multimedia come alive
JOAN ENGEBRETSON
Like the space program in the 1960s, the Internet has become one of the key drivers of today's technology development. Ironically, neither the space program nor the Internet began with the aim of making a profit. But both initiatives ultimately generated huge monetary gains by establishing the framework for a wealth of creative applications.
Many have dismissed the pet projects of academic visionaries as pie-in-the-sky puttering--only to find a few years later that they can't live without the commercial products that resulted. Who would have guessed that the Mosaic browser--originally developed in an academic setting--would mutate into the Internet's killer application?
When the Internet changed from a non-profit entity into a commercial one in the early 1990s, it didn't take long for members of the academic and research community to seek out new venues for their forward-thinking research. Today, numerous regional academic consortiums are operating what have come to be known as next generation Internet networks. Linking these high-tech islands together are two national IP backbones dedicated to the academic and research community--Very High Performance Backbone Network Service (vBNS) and Project Abilene.
VBNS got its start through the National Science Foundation, the same organization that funded the original Internet backbone. As commercial Internet backbones appeared in the early 1990s, the NSF realized that it was "no longer necessary or appropriate for [it] to be in that game," says Rick Wilder, director of Advanced Internet Engineering for MCI WorldCom, which operates vBNS. "What the NSF saw as its next role was to build a network to support the research it was funding in supercomputing."
MCI, prior to its merger with WorldCom, won the contract to build that network in 1994, turning it up in 1995. Although the NSF has funded much of the project, Wilder says MCI chipped in its own laboratory research and engineering services.
VBNS now connects three supercomputing centers and numerous research and academic institutions. A typical supercomputing project is a weather study that involves plotting huge amounts of data on maps and animating those graphics. The vBNS lets users outside of the supercomputing centers apply that data in their work.
The service also has begun to target smaller universities that may want lower speed access lines, Wilder says. NSF funding for vBNS ends in April 2000, but MCI WorldCom expects to continue to operate the network for the foreseeable future, he says.
Project Abilene, which has been in operation since early this year, has its roots in the Internet 2 consortium of universities. The consortium formed in 1996 with the goal of developing applications for the next generation Internet, along with the infrastructure to support those applications, says Doug Van Houweling, president and CEO of the University Corporation for Advanced Internet Development, the non-profit organization that leads Internet 2 efforts.
"We're in a position to do large-scale multinetwork tests of protocols like DiffServ and to discover things that would not be discernible in a laboratory," Van Houweling says.
The Internet 2 group now has 160 university and 55 corporate members, some of which are connected to vBNS and the Abilene network.
Qwest Communications operates the Abilene backbone and has loaned the fiber to support the project for three years, with an extension option. Nortel Networks loaned transport equipment, and Cisco Systems loaned routers for the backbone network. Corporate participants are lured by the prospect of what they can learn by working with networking gurus from the academic community.
"The academic use of technology is often indicative of future markets," says Michael Turzanski, deputy director for advanced Internet initiatives for Cisco. "We like to work with the best and brightest in the field, from academic and corporate environments."
Cisco already has used its participation in Internet 2 to develop new software for its routers, he adds.
Internet 2 corporate and academic members meet regularly in work groups devoted to specific development areas, including quality of service (QOS), multicasting and IPv6--the next version of the Internet protocol (see sidebar this page). Another key initiative is to ensure high performance levels across regional academic networks and individual campus networks to which Abilene is linked.
"We're beginning to focus on the interaction of all these networks so that a faculty member at one university, connected at 100 Mb/s, would be able to connect at a large fraction of that speed with a colleague at another university as a routine matter," Van Houweling says.
Blue sky applications
Some have questioned just how advanced the vBNS or Abilene networks are. "We're pedaling as fast as we can," jokes an executive at one of the commercial Internet backbones.
Abilene operates at OC-48, and vBNS will operate at that rate by mid-2000. But that's no faster than some of the commercial Internet backbones, including UUNet's and Sprint's.
Project Abilene is helping to pioneer packet over Sonet, and vBNS also plans to migrate to that technology by the middle of next year--but the next generation networks are not alone in using that architecture, either. GTE Internetworking, Frontier Global Center and others already are taking the same approach.
What distinguishes Abilene and vBNS is that both networks have relatively few users when compared with commercial Internet backbones, and each individual organization connected to one of these networks has a lot of bandwidth at its disposal.
That extra bandwidth is crucial to many of the applications that are emerging on these next generation Internets--such as Tele-Immersion, an initiative of the central laboratories unit of Internet 2. Tele-Immersion developers are working on transmitting three-dimensional images in real time across great distances. The project is headed by Jaron Lanier, who originally got involved with 3-D imaging as a pioneer of virtual reality applications. Now the lead scientist for the Tele-Immersion Initiative, Lanier hopes to use his 3-D expertise to enhance the collaborative work experience.
Tele-Immersion developers already have a prototype of what they call a Tele-cubicle. Inside a Tele-cubicle, participants can see 3-D representations of people located in another Tele-cubicle thousands of miles away. They can even hand one another representations of three-dimensional objects, although they must rely on visual cues to experience how these objects feel. "You can touch it, but you can't [literally] feel it," Lanier says. Practical applications for the technology could include computer-aided design and remote diagnoses by radiologists.
Key to the project are what Lanier calls environmental scanners, located inside each Tele-cubicle. Multiple scanners collect visual and audio information, transforming it into a 3-D representation for transmission across the next generation Internet. Developers are working on compression methods, but currently, a single Tele-Immersion application can consume OC-12 rate bandwidth.
"The thing that takes up bandwidth is the environmental scan," Lanier says.
The application could use less bandwidth if, say, only crucial objects were transmitted in 3-D. Developers will study how important it is to see one's colleagues in three dimensions, Lanier says. "My guess is that it does matter."
Another application that requires the ample bandwidth of the next generation Internet is Video Space, which is being developed by the International Center for Advanced Internet Research at Northwestern University in Evanston, Ill. Video Space aims to help users get more value out of video content on the Internet by providing a multimedia portal. Like the text and graphic portals available on the conventional Internet, the multimedia portal should help users find the content they seek. Users also will get help in locating a particular segment within a video, says iCAIR Director Joe Mambretti.
Developers are working on a method of storyboarding videos or creating a series of still images that summarize the content. Users then would be able to click on a specific still image and advance to the corresponding position in the actual video.
Supporting the multimedia portal is a video jukebox that can handle thousands of simultaneous multimedia streams. The main challenges facing developers are resolving multicasting issues and ensuring real-time, jitter-free delivery through QOS mechanisms, Mambretti says.
The technology iCAIR is developing to support Video Space should have commercial appeal. "Broadcast video in the future won't be sustainable as a technology," he says. "Smart broadcasters are looking at this."
Anticipating commercial interest in its work, iCAIR hopes to avoid what one might call "Ivory Tower-ism." "We want-ed to get away from the traditional model of researchers being totally isolated and throwing the technology over the wall where it gets shrink-wrapped," Mambretti says.
Instead, corporations can pay for the development of a project prototype or for the use of the technology for their own purposes.
Northwestern University and Internet 2 provided funding for iCAIR and Video Space. Ameritech donated fiber, Cisco donated middleware and IBM provided the supercomputer that supports the Video Space project. Participants also provided human resources. IBM developers have been heavily involved in helping to create indexing and storyboarding functionality, Mambretti says.
IBM sees more opportunities in developing software to support next generation Internet applications, says Michael Nelson, project director for Internet technology at IBM. Most next generation Internet activity "has focused on bandwidth, but bandwidth alone doesn't get you very far," Nelson says. "You have to have middleware and applications."
Infrastructure impact
VBNS and Abilene--along with the other academic and research networks to which they are connected--constitute a closed IP network that operates independently of the conventional Internet (see sidebar). Although no one has given a name to the conglomerate network, virtually anyone connected to either vBNS or Abilene can reach anyone else connected to the other backbone.
The two backbone networks peer with each other in New York and Chicago, where they also peer with some international research networks. But there is no access to the regular Internet through Abilene or vBNS.
Recognizing an opportunity, MCI WorldCom has tailored an offering to organizations that want connectivity to the next generation and conventional Internet networks. It uses a single access line to reach vBNS and UUNet, MCI WorldCom's Internet backbone.
The vBNS network has 12 points of presence (POPs) and three satellite nodes, in addition to the three computing centers, Wilder says. The network currently uses packet over ATM, but that soon will change. At the time vBNS was built, ATM was the only way to do high-speed IP, he says. "But it's getting harder and harder to build ATM interfaces on routers that will keep up. There are lots of economic and technical reasons to believe IP over Sonet will be better than the more complex IP over ATM."
Because it got started later, Qwest was able to adopt a packet-over-Sonet design for the Abilene network from its inception.
"Since we are creating the next generation Internet, we don't want other technologies to substitute for what IP ought to be doing," says Guy Cook, vice president of Internet services for Qwest. "We want to make it as pure and powerful an IP network as we can."
The use of gigaPOPs also distinguishes Project Abilene. GigaPOPs, through which regional academic consortiums access the Abilene network, sometimes resemble conventional POPs. But rather than being operated by Qwest, they are operated by the university consortiums. Some regional consortiums operate distributed gigaPOPs, which eliminate the need for each individual university to run an access line to a central location.
GigaPOPs exist for three reasons, says Cisco's Turzanski. They enable universities to reduce their costs by aggregating traffic, they provide a direct connection to other universities in the regional consortium, and they enable members to share intellectual capital, such as QOS research.
Currently, there are approximately 20 gigaPOPs on the Abilene backbone. In some cases, member universities also access the vBNS backbone through the gigaPOP.
The move toward mainstream
The next generation network operators hope that the applications developed on their closed networks ultimately will be available to the rest of the world. And like conventional Internet backbone providers, they are focusing closely on how they will ensure QOS. Although many conventional Internet providers plan to use a combination of DiffServ and multiprotocol label switching, both vBNS and Abilene will rely primarily on DiffServ.
The Internet 2 consortium held a meeting regarding QOS about a year ago, says UCAID's Van Houweling. "We brought together the best people from around the country, and they formed a consensus that we should start with DiffServ," he says.
MCI WorldCom has been testing a router in the vBNS network from Juniper Networks that supports differentiated QOS, Wilder says. The service provider plans to use it to offer three service levels. These include a best-effort offering typical of IP networks, a premium level for ongoing applications that require bandwidth and latency guarantees and a service based on the resource reservation protocol for users that need premium service for a specific time period.
MCI WorldCom plans to transfer what it has learned with vBNS to its UUNet backbone, Wilder says. Essentially, vBNS served as the proving ground for the Juniper router, which the service provider now plans to use within UUNet.
Van Houweling hopes that applications and networking technology developed on the next generation Internet also will migrate to the mainstream Internet. He points out that only 10% to 20% of conventional Internet traffic that universities generate goes to other universities.
"Most of it goes to the rest of the world," he says. "It's not adequate to have an Internet among ourselves that can support advanced applications. We want to transfer whatever we do to private industry."
If you thought the growth of the Internet was staggering before, just wait until some of these high-bandwidth next generation applications go mainstream.
--------------------------------------------------------------------------------
Higher tech distance learning
Steven C. Myers, interim associate vice president for information services at the University of Akron in Ohio is involved in the Internet 2 consortium of universities because he hopes the group's Abilene network and similar pure IP networks eventually will support what he calls "collaborative classrooms."
These classrooms, which one might call state-of-the-art distance learning, include a sophisticated videoconferencing system, an electronic whiteboard that runs NetMeeting and other high-tech equipment aimed at enhancing the learning experience. Classrooms also can share data and other media, such as the Internet or real-time images, from document cameras.
"It's not just videoconferencing. It's a true collaborative environment," Myers says.
Unlike conventional videoconferencing systems, the equipment used in collaborative classrooms allows participants at each end of the connection to speak and hear each other and share data simultaneously. Those frustrated with the traditional videoconferencing experience invariably say, "This is nothing like we're used to," Myers says.
The setup also is different because participants do not see themselves on screen. "You should never show [people] themselves because it's not natural," Myers says.
The University of Akron operates an OC-12 ATM network that interconnects 14 collaborative classrooms at the university and at regional high schools and two-year colleges. The system, dubbed UADILES (for University of Akron Distance Learning System), enables schools to offer innovative programs such as a Spanish class that links suburban and inner city classrooms. The program provides suburban students with an opportunity to converse with some of the native Spanish speakers who attend the inner city school.
Bright high school students also have the opportunity to take college courses remotely. "Students don't have to, at the age of 15 or 16 or 17, come onto the university campus," Myers says.
He attributes the high quality of collaborative classroom videoconferencing to the MPEG 2-over-ATM standard, a standard not supported by pure IP networks like Abilene (see related sidebar). The University of Akron is undertaking interoperability tests to confirm that UADILES can interoperate with IP-based videoconferencing systems--although sound and picture quality drops to the lowest common denominator of the IP-based system. Eventually, quality of service mechanisms such as DiffServ and MPLS may enable IP to support equivalent quality, but no videoconferencing vendors have built equipment to those standards yet.
An outgrowth of the University of Akron's pioneering work with MPEG 2 is its role as facilitator and founder of the MPEG 2 Forum, which includes all manufacturers of MPEG 2 videoconferencing equipment. Until now, the forum has focused on the interoperability between different manufacturers' equipment, but Myers expects interoperability of ATM and IP videoconferencing systems to become the next hot issue.
--Joan Engebretson |
