Internet 2 and the future of US research networks

By: Scott Bradner

In the beginning Internet 2 (www.internet2.edu) was based on complaining. Many US researchers had been used to the high performance, reliable connectivity provided by the government supported National Science Foundation (NSF) network -- NSFnet. But in the years since the government got out of the business of providing general backbone network connectivity Internet service has become progressively more expensive, while at the same time the quality of the service provided degraded. Last year a group of campus networking managers got together in Monterey California to talk over the future prospects for the support of network-based research in the US. Their discussions led to an effort in FARNET ( www.farnet.org, an organization started by the old NSF-sponsored regional networks) and finally EDUCOM (www.educom.edu, a non-profit organization supporting the use of information technology in education) to understand what the actual future requirements for data networking are in the US higher education community. But these efforts proved to be hard to bring to a useful conclusion. The people working on the issue got progressively more confused as to what was needed.

There was little useful progress in this effort until a meeting in Colorado Springs, Colorado in August 1996. The first half of this meeting was also very confused with the 100 or so people in the meeting failing to come to any understanding. At the end of the 1st day a few of us technical people went into a separate room and finally figured out that there were three separate sets of concerns being expressed. First, there were the people who felt that the commodity Internet had become too expensive for the higher educational community and it was of too poor a quality to be useful. Second, some of the attendees were worried that there existed no network platform for the development of new applications technologies, applications that were going to need a network that supported some form of quality of service controls. Finally, there were a group of people trying to figure out how to enable distance learning by providing high-quality real-time video interaction to thousands of locations in small towns all over America. Clearly the first and last of these issues were not ones that the US higher education community can "fix" on its own. The higher educational community can not build a private "Intranet" for higher educational institutions to fix the first problem because only a small part of the Internet communications of any Institution goes to other higher educational institutions, as much as 90% goes to the commodity Internet. The higher educational community also can not install the network infrastructure required to enable distance learning to small towns, it is just too costly. So these issues have to be off the table. But the second problem has a chance of being approachable. This realization helped focus the discussions in a useful way.

After that meeting I was asked to form a small technical team and come up with a proposal for an architecture for an Internet 2 that would provide a platform for the development of the next generation of network applications, applications which would then be migrated to the regular Internet when the QoS functions became available. Our team came up with such an architecture and we presented it to the chief information officers of about 50 US higher educational research institutions in early October 1996. That group voted to form an Internet 2 organization to coordinate application development efforts and to implement the proposed architecture. The architecture includes local or regional very high speed interconnections between higher educational institutions called gigapops. Gigapops are service interconnection points where the networks of the institutions can exchange traffic with each other or with commodity Internet providers, with research network backbones such as the NSF vBNS, and even with providers of other telecommunications services. The gigapops are to be interconnected with high-speed, QoS enabled networks. The NSF vBNS has been proposed as the initial inter-gigapop interconnection network. At that point things were quite clear. The mission of Internet 2 was clear and the fact that it would be supported by the individual institutions with little or no governmental funds was also clear. But that did not last.

Since then things have gotten confusing again. The Clinton administration announced the Next Generation Internet(NGI), which mentions Internet 2 as a way to provide new high speed connections for one hundred or so higher educational institutions in the same breath as mentioning $100M in funding for NGI. This has caused many schools to see the Internet 2 effort as a way to resurrect the cheap, government supported, high speed Internet connections that went away with the close-out of the NSFnet even though this is not the case. There is some federal funding available for connections to the vBNS but the amount of money is significantly less than the actual cost of the connections.

Meanwhile the Internet 2 organization is proceeding in spite of the confusion and has developed partnerships with a number of major US corporations including Advanced Network & Services, Bay Networks, Cisco Systems FORE Systems, IBM Corporation, and Newbridge Networks, and is proceeding to figure out the actual requirements for the Internet 2 network. Over a dozen gigapops are under active development around the US and explorations are underway to see how Internet 2 can interoperate with international research networks. Connections should be up in a few months between the first few gigapops. QoS enabled interconnections are expected next year with a number of demonstration applications becoming available over the next year or so.

In an ideal world the QoS network functions will be perfected on Internet 2 then moved over to the regular Internet, at which point Internet 2 will go on to work on other advanced pre-competitive technologies--always keeping one step ahead of what is available in the commodity Internet. It will be fun to see if this can happen.