NCSA's New and Upgraded Computational Resources

NCSA installs CRAY SUPERSERVER 6400

In late April, Cray Research Inc. and NCSA announced an evaluation agreement that led to the installation of a CRAY SUPERSERVER 6400 (CS6400) system at the Center. The CS6400 data server is the newest enhancement to NCSA's computational resources for the national research community.

NCSA's CRAY SUPERSERVER 6400 (Photo by Wilmer Zehr)

According to NCSA Director Larry Smarr, the Center is responding to a national need among U.S. businesses for new types of HPC applications. "We see NCSA as a vital commercial application testbed for U.S. businesses to use in prototyping solutions that can be transferred to their own computing environments," Smarr says. "The CS6400 will enable NCSA to continue to provide a leadership role in scalable client-server solutions for business applications. We hope to have further discussions with Cray Research regarding the possibility of team efforts with NCSA in this area."

"The CS6400 has already captured commercial customers in the insurance, travel and tourism, financial services, general manufacturing, and managed health care industries. [It] will be applied to very large databases in the multiple terabytes range," states Robert Ewald, Cray Research president and chief operating officer. "We are pleased to bring our business computing expertise to NCSA and look forward to working with NCSA on this critically important initiative."

NCSA is building a network of business partners, hardware and software vendors, and academic researchers to make scalable applications available that tap the powerful resources of supercomputers and data servers like the CS6400 system. NCSA will use the CS6400 data server for a variety of applications including geographical information systems (GIS), intelligent algorithms, and parallel structured queries on large databases.

GIS provides environmental and demographic data for use in business, education, and commerce. "The CS6400 is ideal for this type of application because it can easily handle high data transfer rates and processing while managing heavy storage burdens," Smarr adds.

In addition, the CS6400 will be used to augment traditional methods of data analysis with techniques developed in the artificial intelligence community, including machine learning and genetic algorithms. "These techniques are referred to as intelligent algorithms," Smarr says. "Our goal is to discover, within very large quantities of data, patterns that are of interest to the user."

Although the goal of fully automating the process of knowledge discovery has yet to be realized, numerous algorithms have been developed with which exploratory work can begin using real-world datasets on today's HPC architectures. NCSA believes there are major opportunities to work with companies who wish to explore these techniques on their own data.

NCSA will be exploring leading-edge parallel database products- such as those provided by Oracle Corp., Informix, and Sybase Inc.- to the multigigabyte databases being generated by NCSA's industrial partners. The parallel architecture of the CS6400 data server, along with parallel database technology, will enable decision support systems and other query-intensive applications to help businesses find new competitive advantages within their databases.

The CS6400 system at NCSA has 10 SuperSPARC Plus microprocessors, 1 gigabyte of central memory, and 40 gigabytes of online storage. It runs the Solaris 2.4 operating environment, which has been parallelized. The system is connected to NCSA's research environment via HIPPI, FDDI, and Ethernet.

Other new and upgraded computational resources at NCSA are listed below.


access / Summer 1995 / NCSA