NETWORKS

By Steve Fisher, Editor In Chief -- Networking is of course of paramount importance to supercomputing applications. So, when Supercomputing Online wanted to do a story focusing on networking (particularly at the recent SC show) we turned to one of the industry's leaders…Qwest. The following interview is with Wes Kaplow, Chief Technology Officer, Qwest Government Services. Supercomputing: Qwest participated in the High Performance Bandwidth Challenge at SC2001 did it not? How did things go? KAPLOW: This is the second year that Qwest sponsored the Bandwidth Challenge. Following in the spirit of last year, there was a wide range of applications that competed for the prize. In general, the applications appeared to be more finished, with an eye toward the use of the network either fostering new types of collaboration, or providing access to a grid of supercomputing resources. Applications ranged from the use of networking to provide a virtual intercontinental collaboration of musicians and dancers from Brazil, Denver, Minneapolis, and Florida, to the real-time remote control of an electron microscope and coordinated use of supercomputing resources to create three-dimensional reconstructions. The Bandwidth Challenge has provided a focus for the innovative use of high-performance networking, and Qwest will continue to support and refine the goals of the award to foster even more challenging uses of gigabit and beyond networking. Supercomputing: Networking obviously plays a huge role in supercomputing applications. Please tell us about some of the work you (Qwest) have done with some well known supercomputing providers, institutions etc. KAPLOW: Qwest is the provider of network services for the Department of Energy's Energy Sciences Network. Qwest is both the provider of backbone bandwidth and currently seven OC-12 and above ESnet locations. Qwest has built fiber into ANL (Argonne), ORNL (Oakridge), SLAC (Stanford), LBNL (Lawrence Berkeley), LANL (Los Alamos), SNLA (Sandia Albuquerque), and NERSC (DoE Supercomputing facility) with LLNL (Lawrence Livermore), SNLL (Sandia Livermore) on the way, and possibly FNL (Fermi) to follow. These locations, such as NERSC, are some of the largest supercomputing sites in the world. Qwest is part of the NSF TeraGrid program (which includes ANL, NCSA, CalTech, and the SDSC) and is working with the Principal Investigators to create a network and supercomputing complex that will have an initial network capability of 40 Gbps from the locations in Illinois to the locations in California. These are some of the nation's premier supercomputing locations and research institutions, and Qwest is working with the team understand the unique demands of connecting supercomputers to multi-gigabit networks, and to develop a scalable approach for adding additional supercomputing centers to this network. Qwest also provides OC-48 IP connections to the PSC (Pittsburgh Supercomputing Center) as part of the DARPA Supernet program, and provides the infrastructure and works very closely with UCAID/Internet2 serving over 180 universities, providing broad access to some of the nation's most capable supercomputing centers. Supercomputing: Please tell the readers about what you were highlighting at Qwest's SC2001 booth. KAPLOW: Our booth theme was: Bigger Science, Smaller World, Network by Qwest. These items are represented by our commitment to the advanced networking and scientific community. This includes the DoE ESnet, UCAID/Abilene network, NASA's NREN, and NSF's DTF (a.k.a. TeraGrid) program. These networks are used to enable scientific collaboration (the smaller world), and the development of new geographically diverse supercomputing complexes to attack every larger problems (the bigger science). Qwest also highlighted our metropolitan networks in 25 cities that have been used to break the "last mile" bottleneck of getting cost effective high-capacity networking from the backbone into customer locations. Supercomputing: SC Global was something very interesting at this year's show as far as grid technology is concerned. What did you think of it? KAPLOW: It is important to think globally and attempt applications that challenge the capabilities of today's networks. As scientific research becomes increasingly more collaborative and international, we must also keep in mind the difficulties of working in countries all around the world and continue to explore new collaborative tools. Supercomputing: What were you most impressed by at the show? In general how do you think it compared to the shows of the past? KAPLOW: What is remarkable is the ability to create a supercomputing center connected with super-networking capability so quickly. The coordination required to create the SCinet at the show and connect it to the major research networks in the United States is tremendous, and this year was no exception with several more gigabits per second of bandwidth capability to the world as compared to last year. In general, the level of refinement in applications demonstrated this year was improved over last year. More focus seemed to be placed on using networking to improve not only the computation time for a result, but also to improve the computer to researcher interface. Supercomputing: Is there anything you'd like to add? KAPLOW: Qwest and I have been impressed by sense of vision and drive of the government and scientific community, many of which were represented at SC2001. These individuals and organizations are challenging the limits of machines and networks, undaunted by what is possible today, driving to create the computing infrastructure required to attack scientific problems of national significance. They are also showing leadership by demonstrating working prototypes of software and computing environments that will certainly be replicated for commercial applications. ---------- Supercomputing Online thanks Wes Kaplow for his time and insights. ----------