Systems and Services

Thanks to the careful planning and preparation of Adrian Wong, Clayton Bagwell, Jonathan Carter, Brent Draney, Mike Welcome, and others, NERSC's T3E-900 was placed into full production service within one week of its installation and acceptance. Their work included software installation, operating system configuration, system tuning and testing, account installation, queue scheduling, and operational support and monitoring.

Like any other commercial venture, high performance computing has its share of exaggerations. In 1998 we heard various vendors make contradictory claims of who built the world's fastest computers and who dominated the market.

Experienced users viewed these claims with skepticism, knowing that most of them were based on theoretical peak performance and manipulation of statistics, not real-world results.

At NERSC, we define high performance in practical, not theoretical, terms. High performance to us means providing the most productive and reliable computing, storage, and networking systems, while consistently offering our clients timely services, innovative assistance, and convenient training.

For example, at the beginning of FY98, we promised to deliver to scientists 2.75 million T3E hours and 860,000 normalized CRUs for the PVP systems. (A CRU is a CPU hour x a particular system's performance factor x a variable priority. Normalized CRUs set the priority to 1). We were proud to deliver more computing time than we promised, as shown in the table below.

This past year's improvements in our systems and services are described throughout this section.