THE I-WAY AND THE FUTURE OF DISTRIBUTED SUPERCOMPUTING

Guest Editorial by Rick Stevens, CRPC Executive Committee Member

We are fortunate this year at SUPERCOMPUTING '95 to witness the convergence of Internetworking, high-performance computing, and advanced visualization environments. The Information Wide Area Year (I-WAY), an experimental high-performance network that will link the country's most powerful supercomputers with advanced visualization technology, is a demonstration of both the best of the computer science community's "volunteerism" and new ideas born of difficult tasks.

The I-WAY is an attempt to create a national testbed for exploring wide area visual supercomputing and an effort to experimentally determine the software architecture of such a testbed. Unlike the gigabit testbeds, the I-WAY is nationwide in scope and based on ATM technology. By connecting approximately 30 supercomputer sites into a single computing resource, the I-WAY is pushing the envelope of distributed supercomputing and the boundaries of collaboration. As in any large-scale, largely volunteer activity, the social and organizational challenges here are often greater than the technical ones.

The I-WAY is being driven by more than 60 advanced applications. The applications range from real-time visualization of clouds to robotic telepresence to interactive, distributed, multi-user virtual reality interfaces for computational fluid dynamics and finite element models. Many of these applications are attempting for the first time to run a single job on "meta" computers consisting of several parallel supercomputers interconnected via the I-WAY network.

By combining computational science simulations with a powerful virtual reality interface, the I-WAY is enabling the development of a new class of supercomputing applications. Because users directly interact with the simulation by pointing, clicking, and dragging in three-space, they are free to concentrate on understanding the dynamics of the scientific problem and not the computer interface. At the same time, I-WAY applications will be pushing the limits on network bandwidth and latency. By putting the user "in the loop," the real-time response of the network becomes a major issue. Much work remains to be done to understand the effectiveness of ATM networks as the basis for the next generation of TCP/IP internetworks. The I-WAY is one effort to explore this territory, and we are certain that there will be many more in the future.

In the year of the Internet and World Wide Web, it is appropriate that the SUPERCOMPUTING '95 conference has both the I-WAY project and the largest networking budget of any previous conference. That the high-performance computing community is still driving the future of the Internet with advanced applications may surprise many outside the scientific community, but should not surprise those truly involved in the field. The refocusing of high-performance computing and the emerging pervasiveness of the Internet will provide the HPCC community with considerable opportunities for experimenting with new and innovative ideas. We should truly celebrate the successes of HPCC and enthusiastically encourage the pursuit of the next generation of tools, applications, and paradigms for building the global information infrastructure.

For more information on the I-WAY, see http://www.iway.org/ .

Table of Contents