Jeo W., I still remember someone asking ANCR's Cal Nelson about the impact that Internet2 could have applications for SANs and FC.............It was a log ( 10X) effect.
Internet Two Too Much -- They're At It Again, Weaving Away To Create A New Net. Let's Hope It Doesn't Get Too Sticky.
Kate Gerwig
It's not politically correct to question the value of Internet2, although it certainly has its detractors. It is perfectly acceptable, however, to ask what it is. Getting a straight answer is another matter, as even its proponents admit.
The problem is that the Internet2 initiative is a lot of things. It is both a concept and a set of competing and cooperative private networks. The picture is further clouded because the initiative is promoted by different interests-academic, government and industrial. What is clear is that the newest Net surfaced in 1995 after the original Internet went commercial, causing great dismay for the many in the research community who viewed it as their private playground.
The latest offering's aim is clear enough. It is seen as a private research domain for developing advanced Internet applications and technologies that should eventually find their way into commercial circles. Beyond this, "Internet2" has become the generic handle for the two national research-oriented networks. Abilene was built by official Internet2 members beginning in 1997, with donations from Qwest Communications International Inc. (Denver), Cisco Systems Inc. (Santa Clara, Calif.) and Nortel Networks Corp. The other is the very high performance backbone network service (VBNS), which was commissioned by the original Internet's founder, the National Science Foundation (NSF), in 1995 and built and run by MCI WorldCom Inc.
Although they have the same general purpose, these two networks were separately created and funded. While they have different memberships, they are now interconnected to each other and to other government and research networks around the world.
"It got messy because the names started getting all jumbled. People started thinking that Internet2 meant connections to Abilene and that Abilene was the NSF's network. The community we serve is totally confused, and I'm sure the public and the press are too," says Charles Lee, an executive manager at MCI WorldCom, which is marketing its new VBNS+ service to government agencies, research institutions and universities that want access to a private Internet protocol (IP) network that doesn't carry commercial traffic.
Confusion is only one of the challenges facing Internet2. It's far more daunting to figure out whether the initiative is worth the effort and the tens, if not hundreds, of millions of dollars being thrown at it. Opinions abound on the matter, although a lot of people will only voice them off the record.
No one really thinks Internet2 will bring about the business revolution that its namesake produced. There are some who acknowledge that some of the applications research now being conducted on Internet2 could have interesting uses in the commercial sector, but it's debatable whether those applications couldn't be built off the existing Net.
These questions reflect many of the central concerns over Internet2's value. They probably also explain why the latest initiative doesn't enjoy the same industrial support as the original. Industrial support abounds: Companies like Cisco, IBM, MCI WorldCom, Qwest and Nortel are heavily subsidizing Internet2. But other firms that played leading roles in the development of the original Internet, like GTE Internetworking (formerly BBN, Burlington, Mass.), are barely involved. They're not abandoning next-generation Internet applications, but many are using their own private test networks and relying on the in-house expertise gained in creating the original Internet.
Concerns over Internet2's worth were underscored recently when Vint Cerf, the man credited with creating IP, questioned whether its research efforts can get far enough ahead of the industry's own efforts to make it worthwhile. His concerns no doubt struck a particularly deep chord, given his position as senior vice president for Internet architecture and technology at MCI WorldCom, which supplies one of the next-generation Internet's high-speed backbones.
Some critics have even gone so far as to call Internet2 a political maneuver by the White House and certain government agencies to justify sending money to educate graduate students. "In my mind, there is no value created by Internet2. I'd really rather just give money to the colleges to give the grad students. We would not participate, ever," says PSINet Inc. (Herndon, Va.) chairman and CEO Bill Schrader, who broke with the research community to found the first commercial ISP a decade ago.
Critics say Internet2's payback is too low and too slow. They claim that some universities are already working on unrealistically complex 3-D virtual reality and telemedicine applications. Researchers are also working on many of the same issues as the commercial sector, including quality of service (QoS), multicasting, caching and distributed content storage. "Internet2 attempts to follow along the same path as the original NSF network program, bringing esoteric information transport technologies out of the laboratory and toward production use. But the NSF did its job incredibly well, and we now need to allow commercial organizations to develop new techniques and compete in the global marketplace," says Joel Maloff, president of Maloff Group International (Ann Arbor, Mich.).
In terms of networking, Internet2 doesn't do anything that the commercial Internet doesn't do, says Dennis Fazio, former executive director of MRNet, once one of the regional NSF networks. But when Internet2 was developed, there was a subtext that the original Internet wasn't as good as Internet2. The truth is, when it comes to usage, the two are different animals entirely.
So what is the point of Internet2? The uncynical view is that it will give computer scientists a place to develop advanced technologies that eventually will reach the commercial Internet-just like the first time. Academics claim a private research network is essential. The original Internet isn't applicable for advanced research because of its high level of production traffic and uneven bandwidth availability, Fazio says. "There is an expectation of reliability on the commercial Internet that isn't necessary on a research backbone. If it's broke, that's fine-that's what happens when you do research. You stretch the limits. You don't want to do that on the commercial Internet."
The more cynical viewpoint stems from lingering resentment among academics over the original Internet, which they helped create, only to see it turned over to companies that made money from their developments. This view speculates that a hidden university agenda exists to rebuild a high-speed private IP playground that would replace the old one. Only this time, the theory goes, academia will keep control.
Officials at Internet2, which controls the Abilene network, seem to provide reasons for supporting both sides. "We don't see turning any network services over, the way NSFNet was privatized in 1995," says Ted Hanss, director of applications development for Internet2. "We're very focused on technology transfer through our corporate partners. Working on things like QoS and having Cisco and 3Com build it into their products will raise the capability of the broader commodity Internet."
The debate over its worth notwithstanding, one thing is for sure: The buzz that surrounds the commercial Internet on a daily basis is nowhere to be found around Internet2. Its relative obscurity and confusing patronage no doubt limits interest. It also may be simply too early: Much of the work to date has involved just getting the two main research networks running at high enough speeds to handle the most advanced applications (see "A Tale of Two Nets"). Only now are the research networks beginning to see spikes in traffic from advanced research applications, says Guy Cook, vice president of advanced Internet services at Qwest.
The other factor limiting widespread interest is reality. The "commodity" Internet-as the original is sometimes called in research-speak-was one of a kind, like the development of the mainframe computer and then the personal computer. Even Internet2's proponents acknowledge that the possibility of Internet2 developments revolutionizing society-again-is slim. "The Internet is going gangbusters, and everybody knows it," says Cook, who oversees much of Qwest's involvement with the Abilene network, the OC-48 (2.4 Gbit/s) Internet backbone that Qwest donated to the Internet2 project for five years. "Internet2 should be broadly defined as a variety of different networking efforts in national labs, the federal government and universities that have research and development as their main objectives."
Payoffs may come, but they're likely to be years away. "Internet2 is using leading-edge applications as an incubator for new approaches that may serve the commercial marketplace in years to come," says Maloff. Internet2 researchers, for example, are working with interactive virtual reality and advanced physics applications that eventually could be used by car manufacturers. Virtual applications would enable 24-hour production cycles where a design team in one country could hand off its work to a team in another country. What Internet2 researchers aren't working on are e-commerce and Internet-enabled traditional business processes, the sorts of things that might create a buzz in the business world.
The research community also has very different networking needs than the commercial sector. The commercial Internet doesn't have the capacity and QoS capabilities to experiment with advanced apps today, says Hanss, "but we're trying to get applications developed in anticipation of that happening. Then we will move on to Internet3 challenges, the next generation of advanced applications that will exploit things like voice networking."
One major difference between Internet2 and its forebear is that commercial industry partners have subsidized its efforts from the beginning. And despite all the concerns about payouts and academic control, the 20 corporate sponsors may ultimately get more out of this than the 160 participating research institutions. Internet2 gives network providers, and hardware and software vendors, a place to try out new Internet technologies without exposing commercial customers to the starts and stops of experimental networking research.
Internet2 is a real-life, real-time test bed for technologies that will become a critical part of the next generation of the Internet, says Rich Wall, program director of Internet technology at IBM, one of Internet2's original corporate partners. "Internet2 has gotten caught in some misnomers that make it sound like a repeat of Jaws 1, 2 and 3," Wall says. Its potential involves working on applications that can't operate well in today's Internet because of lack of bandwidth and QoS, which he says is why IBM supports Internet2 and similar research efforts worldwide.
For its part, IBM is looking at applications and middleware that operates between the applications and the network infrastructure to enable things like security, authentication and policy management. Beyond this, its three main areas of interest are advanced video distribution, QoS, and replication, storage and caching.
Video-FC is good at this. Replication, storage and caching-FC is good here too.
Researchers can do experiments that large ISPs aren't willing to do on their own networks, says Hanss. "The commercial Internet is supposed to be reliable in a way that a research backbone doesn't have to be, since its whole purpose is experimentation," he says. And because the research effort has access to more than one Internet backbone, it can test things like cross-network QoS and possibly troubleshoot new protocols before they reach commercial networks.
Some companies are willing to gamble millions on what Internet2 may be able to deliver. Qwest could charge $500 million in the commercial market for the OC-48 network it donated, Cook says. Instead, Qwest thinks of Abilene as a giant playground for experimentation. "We can throw anything we want to try out as an industry over the transom and let the university network managers help us," Cook says. "We don't know how to manage the kinds of high-speed traffic streams that are likely to result in a relatively short period of time."
Internet2 also gives vendors and service providers the chance to get something right before putting it into the market. "You don't have millions of modems connected to this network, so it's optimized to serve a different kind of customer. This is a smaller-scale community, and we can try things out earlier. When something new comes out, it has to be extremely solid, tested and wrung out before it can go into the large-scale infrastructure," says Rick Wilder, MCI WorldCom's director of advanced Internet engineering.
That's the idea. In the best of all worlds, Internet2 should only augment its heavyweight predecessor. "The Internet is there. It can only happen once, and everything else is built on top of that," says Fazio. "That doesn't make Internet2 a worthless project. Some things will migrate out to the great unwashed masses using the commodity Internet."
---
A Tale of Two Nets
Can there ever be too much of a good thing, like too many Internets? Maybe. But that hasn't stopped two major research Internets from setting up shop in the last five years, which has led to confusing and at times competing overtones. The dust had barely settled in 1995 from the National Science Foundation's decision to open the original Internet to commercial interests when the NSF went to work creating another test-bed network. The aim was to connect supercomputing centers in hopes of advancing networking and applications research. The foundation's efforts centered on a five-year deal with MCI WorldCom Inc. to develop this Internet high-speed network, which resulted in the creation of the very high performance backbone network service (VBNS). The NSF helps support VBNS through an annual $10 million grant, and MCI WorldCom sinks at least that much into it each year. The NSF also provides grants and authorizations for institutions to connect to VBNS. A total of 100 are now connected.
Within two years of the inception of VBNS, a group of universities formed the University Corporation for Advanced Internet Development (UCAID). This group serves as an umbrella for academic researchers, hardware and software vendors, and service providers. UCAID dubbed its effort Internet2, giving the initiative virtually the same aim as VBNS.
The big difference is in the backbone. UCAID worked out a five-year deal with Qwest Communications International Inc. (Denver), which was just building its new fiber optic network. Under the arrangement, Qwest donated a national OC-48 (2.4 Gbit/s) network to UCAID called the Abilene network. Cisco Systems Inc. (San Jose, Calif.) provided the routers, and Nortel Networks Corp. provided switches for the research network.
Admission to the Abilene network requires universities and research institutions to pay $40,000 a year for UCAID membership. They must also promise to spend half a million dollars a year in campus infrastructure upgrades so that the network can deliver data to campus desktops at a speed of at least 10 Mbit/s. About 30 universities have connected directly to Abilene, and another 70 have shared access.
In contrast, MCI WorldCom runs an OC-12 (622 Mbit/s) IP over an asynchronous transfer mode (ATM) backbone. It is in the process of upgrading its network to OC-48 and will then operate dual technology backbones-one based on ATM, the other IP over Synchronous Optical Network (Sonet), says Rick Wilder, MCI WorldCom director of advanced Internet engineering.
The two networks, while separate, are interconnected at various points. So are other U.S. government and global research networks, such as ones controlled by NASA and the U.S. Department of Energy.
The future of these private networks is just as murky as the present. It is unclear, for example, what will happen after Qwest's five-year donation of Abilene runs out in 2003. NSF has extended its VBNS contract with MCI WorldCom for three more years, but the $10 million annual subsidy disappears in April. MCI WorldCom is currently selling its VBNS+ commercial service to universities and government organizations, or any other customer who needs access to a private IP research network.
It is possible that the declining corporate and government subsidies could make connecting to either VBNS or Abilene too expensive for academic groups. That could be the fastest way to kill Internet2. |
