American versus European approaches to academic and research networking

Mannheim, 20 June 98 In the European countries, academic and research (A&R) network services have always been organized on a national basis, and were kept separate from the general purpose Internet. Former head of CERN's Computing and Networks division in Geneva, David O. Williams, regards this natural development as a major advantage in comparison with the traumatic privatization experience in the United States. At the Mannheim Supercomputer Seminar '98, he insisted on the careful sustaining of these valuable national research networks. The European problem is at the level of pan-European and transcontinental interconnectivity. We are beaten by the USA if it comes to extensive bandwidth capacities. In the US, Internet congestion, even from and to rich and prestigous universties is a major problem too. American projects, such as Internet2 and the Next Generation Internet constitute promising initiatives to solve network congestion. The European answer consists in the upcoming TEN-155 project. How should we actually deal with insufficient Internet performance?

Statistics show that packet loss during Internet traffic can be dramatic at some times and may exceed 10% when averaged over a whole month. This is due to the regularly occurring bottlenecks in the routers and the lack of sufficient bandwidth. Reason enough to focus on technology, economics and network organization in the future, according to Mr. Williams. In fact, we can distinguish three generations of Internet applications. The first one includes what we are used to, such as e-mail, Web-access, manually initiated file transfer, telnet and X-window. These services are not what one could call interactive and adventurous, they do not suffer too much from congestion. The second generation provides streaming audio and video for individuals, groupware for collaboration at a distance, automated data access and transfer, and shared file systems, requiring a great amount of bandwidth and reliability. Congestion is a problem for these applications.

For the third generation, we really suffer from shortage of bandwidth, since this forthcoming Internet development includes collaboratories and advanced groupware, remote control rooms and virtual reality implementation at a distance. The various technologies needed to offer these services range from the fairly inexpensive optical fibre transmission and switching over electronic signal processing at speeds exceeding 2.5 Gbps, and using multiple wavelengths, to appropriate routers and switches, which frankly display no fundamental limitations. We also evolve towards so-called differentiated services, which may cause the start of a new "religious war", as Mr. Williams points out.

We have to understand that in the United States, the funding of the federal agencies has a completely different source than the funding of the universities,which is done mostly private. In 1994/95, the universities were forced to look for connectivity from a commercial Internet Service Provider (ISP) since the National Science Foundation (NSF) net backbone was being commercialized. They selected MCI because this provider had been the supplier of the NSFnet backbone. Due to the growth of the general purpose Internet, the Internet soon was dealing with a major congestion. This meant the start for the setting up of Internet2, which will be established for production services but not for proper research networking.

Internet2 is managed by the University Corporation for Advanced Internet Development (UCAID) consortium, consisting of a membership of some 200 American universities. The overall idea is to construct a network of GigaPops which is raising severe competition between the different backbone providers. For Europe, it would likewise be useful to invest in EuroPops in order to improve the international connectivity, says Williams. Since mid-April of this year, the three collaborating companies, Qwest, Cisco and Nortel, are building an alternative backbone in the USA, referred to as the Next Generation Internet (NGI). The initiative has been taken by the federal agencies to set up this tree year programme, starting in 1999 and ending by 2002, and which exclusively deals with research purposes. In the end, it should provide the means to organize first rate medical care in the States, among other services. As it is, Internet2 and NGI are separate initiatives which have nothing to do with each other.

In Europe, the TEN-34 project has delivered high-speed connection for academic research networks. The European Commission has been funding this initiative up to 40% while the remaining 60% came from the connected national organizations. TEN-34 will probably be extended to the end of this year and will be succeeded by the TEN-155 project. This network will have to operate at a speed of 155 Mbps. In fact, up till now, there has been more capacity between Europe and the United States than between the European countries themselves.

TEN-155 will be operative for one year and it is expected that after, a programme within the Fifth Framework will take over.

For the European A&R community, it is of vital importance to make politicians aware of the tremendous impact of Internet services and of the need to largely invest in them. In comparison with the United States, a lot of work remains to be done in this area.

You will find more details on this Mannheim Supercomputer presentation at the Web site of David Williams at CERN.

Leslie Versweyveld