May 15 — The 2015 Cray User Group meeting was held in downtown Chicago April 26-30. Around 200 attendees, both Cray customers and users, enjoyed a range of presentations and discussion sessions from tutorials to discussions. NCSA and the Blue Waters project were the local hosts for this year’s event.
NCSA Director Ed Seidel gave the keynote talk about big data and collaboration in supercomputing and highlighted some of the discoveries enabled by the Blue Waters supercomputer, the largest system Cray has ever built and one of the most powerful supercomputers in the world. “HPC is not about doing science faster,” said Seidel. “It is about doing science that can’t be done any other way.”
Seidel says that this is also what draws young people, the future of HPC, into supercomputing. They need it for their science, but they are not always as well-versed as they could be in supercomputing. He put out a call for increased focus on training new and potential users to maximize system capabilities. Later in the conference NCSA’s Scott Lathrop pointed his audience to several HPC training resources like the HPC University and the Virtual School of Computational Science and Engineering. The Blue Waters projects offers internships for undergraduate students, graduate fellowships, and online classes, all to help inspire and train the next generation of computational scientists.
In addition to training, Seidel noted that theory, computation, and instrumentation in many research fields need to merge somehow to realize the full potential of each.
CUG2015 attendees heard from several Cray executives, most notably Pete Ungaro, Cray’s president and CEO. Ungaro talked about the current state and future of Cray on April 28, and then answered questions from customers on April 29.
A running theme of the conference seemed to be that integrated systems—those that can run simulations, analyze the data, and spit out visualizations—are needed for the future of HPC. Specialized systems require that researchers move data between systems at various points in their workflow. This is both time-and resource-intensive.
Cray invited users to send them any other ideas on how the company can facilitate Big Data better.