High-Speed Networking in Europe: Seeking to Equal U.S. Levels

By Norris Parker Smith
Supercomputer Review

November 1989

Stuttgart, West Germany--Where does Europe stand in high-speed networking for supercomputers? Ahead of the United States? What is happening in advanced technology and new standards for high-bandwidth communication?

Up-to-date answers to these questions were provided in a mid-September meeting here of TGV, an informal group of systems managers and users from all over Europe. (The acronym, a play on the term for France's high-speed train system, stands for Telematique Grande Vitesse, translated as High-Speed Data Networking.)

The session was hosted by RUS (Regionales Rechenzentrum der Universitat Stuttgart, or regional computing center of Stuttgart University), which operates a CRAY-2 and has just installed a Convex system. This was the seventh meeting of TGV since its inauguration in March 1987.

The overall picture of networking in Europe is mixed. Europeans, especially the Germans, are moving ahead with some advanced projects, including high-bandwidth (a potential of 140 megabits per second--Mb/sec--up to 1,000 kilometer) systems that should become fully operational within the next few years.

"Catch up with the U.S."

For the present, however, researchers in science and engineering throughout Europe are in roughly the same situation as supercomputer users in North America in the mid-1980s, when networks such as ARPAnet, MFEnet, BITnet, and the early stages of NSFnet, were the only resources available.

Currently the most highly developed international networks in Europe serve specialized research communities, especially high-energy physics and aerospace. Most circuits are 64 kilobits per second (Kb/sec) or lower. Networking for aeronautics includes some links up to 1 Mb/sec. Italy is establishing a 2 Mb/sec national backbone that will be extended north to CERN, the European research center for high-energy physics near Geneva.

According to a report submitted late this spring to the European Economic Commission by RARE, a study group of researchers using supercomputers, scientific communities requiring improved networking include: combustion analysis, astronomy and space physics, fusion research, the automobile industry, and meteorology.

Europe's PTT's, the fragmented monopoly of postal and telecommunications authorities, remain a crucial barrier. RARE called on EC [European Community] to help.

The RARE report noted that " high-speed networking in Europe is considerably hindered, particularly when collaboration and competition with the U.S.A. are taken into account, by the extraordinarily high tariffs for high-speed lines. Today, a 1.5 Mb/sec line in the U.S.A. costs the user the same amount as a 64 Kb/sec international line of the same length in Europe; in general terms the PTT charges here are seven to ten times those in the U.S.A. for the same line capacity."

RARE has proposed a European High-Speed Networking Initiative, to consist of a central backbone called VENUS, regional networks and supporting networks for specific disciplines. According to J. Prevost, chairman of the RARE working group, there would be 20 to 40 nodes around Europe, serving from 200 to 500 sites.

The RARE report to the EC called for a "phased improvement in network speeds, with 2 Mb/sec now, 100 Mb/sec in a few years' time and Gigabit speeds by the end of the century..." RARE noted that the National Science Foundation backbone in the U.S. was upgraded to 1.5 Mb/sec in 1988 and is to be raised to 45 Mb/sec in 1990.

Initial costs for the European initiative were stated as the approximate equivalent of $5 million U.S. dollars annually in the near future, rising later to roughly 20 or 35 million. Prevost expressed work on the initial phase could begin when funding is available, perhaps in 1990.

He told the TGV session that "Europe needs to catch up with the United States."

Germany's fiberoptic Autobahn

West Germany's PTT, the Deutsche Bundespost (DBP) , has placed an opportunity and challenge upon the doorstep of RUS and other elements of Stuttgart University. The DBP is developing what is to become a national fiberoptics network, with capacities up to 140 Mb/sec. This "Forerunner" system is described as a "fiberoptic autobahn."

It is intended for video-teleconferencing and other high-bandwidth requirements and will also be made available for scientific networking. RUS and collaborators at Stuttgart are developing technology that will make it possible to use up to about 100 Mbits of the new system's capacity.

Two projects are being undertaken. The most dramatic extends north about 1,000 kilometers, connecting the computing facilities at Stuttgart with the Alfred Wegner Institute at Bremerhaven, specializing in oceanic and arctic research.

Paul Christ, in charge of networking at RUS, said that the initial demonstration of this link is to take place in February 1990. This project is supported by technology from the American firm, Ultra Network Technologies.

Another fiberoptic line, extended close to 150 kilometers, links the universities of Stuttgart, Karlsruhe, and Kaiserslautern. Christ described the current state of work as "beginning to begin." Several years will be required to achieve full 100 Mb/sec capacity. Intermediate targets include operation at 2 and 10 Mb/sec.

Institutional as well as technological issues must be worked out. The Bundespost is in the midst of deregulation. Cost factors and pricing for high speed networking have yet to be settled. As Christ noted, there "is good will but it takes time."

Gigabit slotted ring at Cambridge

Another project in Europe was described by David Greaves of Cambridge University, where a metropolitan-level network with about 200 kilometers of optical cable has been established.

The network topology is a slighted ring, which Greaves said permitted cheap hardware and low complexity.

University facilities are spread over relatively long distances within the city. Cambridge earlier had built up a number of local networks, with lengths in the hundreds of meters, also slotted rings, design, whose design is called the Cambridge Fast Ring. The new project, built with support from the Olivetti corporation of Italy (which has a facility at Cambridge) , is a backbone network linking the individual Fast Ring local networks.

The potential of the system is in the 1 Gb/sec range. It has been tested at 600 Mb/sec.

An optoelectronic future?

Developments in U.S. as well as European technology were explored during the TGV session. Jon R. Sauer of AT&T, now attached to the Center for Optoelectronic Computing Systems at the University of Colorado at Boulder, described a Gigabit-level project, undertaken in collaboration with the Center of Telecommunications Research at Columbia University, that has received support from NSF.

The project calls for designing and building a "multi-Gb/sec optoelectronic data transport network using self-routing packets in a multi-hop network." It employs custom $2,000 directional coupler switches, made from lithium niobate, which permit the packets to remain in optical form throughout their journey from source to receiver, including switching.

In existing technology, data carried in optical form over fiber must be converted back into the electronic mode for switching. A paper by Sauer said that this technology would permit peak point-to-point fiber bandwidth above 10 Gb/sec, without requiring high-speed electronics or precision synchronization.

Plans call for a lower speed, proof-of-principle four-to-eight node network in two-and-one half years, and a higher speed, larger engineering demonstration with 24 nodes in five years.

Don't call it HSC, But it's just about here

Developments in the new HSC (high-speed channel) standard for high-bandwidth communication were discussed by John Morrison of Los Alamos National Laboratory, which has led HSC development, and Newt Perdue, a vice president of Ultra who had spent the preceding three months in Europe working with RUS and other institutions.

Morrison confirmed that development of the standard is nearly complete, and final definition by ANSI is expected to take place in 1990. It is designed for easy implementation with off-the-shelf parts. It also will be proposed for approval as an international standard.

As the standard approaches maturity, however, a change of name might be necessary. Digital Equipment Corporation (DEC) has been one of about eighty participants working on the standard, but DEC's attorneys have discovered in their database that HSC is a registered trademark for a DEC product called Hierarchical Storage Controller.

Proposals for a new tag include HRC (high-rate channel) HRI (high-rate interface) and a few names that might impinge upon well-known trademarks.

Morrison reported that over 15 HSC-related projects are in active development. IBM has already brought an HSC product to market. Others involved in HSC product development include: Cray Research, Inc., DEC Maximum Strategy, communications firms such as Network Systems, and SCS, as well as Ultra, Sun Microsystems, Tektronix and Thinking Machines.

NSFNET is out front!

Some supporters of an early upgrade of the NSF backbone, above the Gigabit level, have asserted that the Europeans are getting ahead of the United States in telecommunications. The discussions at the TGV session here did not confirm this. Fortunately, the proposal for an upgraded NSF backbone stands on its own merits.

It is a tragedy if scientists and engineers anywhere are handicapped significantly by inadequate access to distant computing facilities. This perspective is much sounder than a perspective of competitiveness. This impression was strengthened by the free exchange of information and enthusiasm for telecommunications progress under all flags demonstrated at TGV's Stuttgart meeting.

Taken from The Link Letter, December 1989, Vol. 2 No. 6.