Data-driven challenges to architectures and systems
Panel Chair: Dr. Chandrika Kamath, Computer Scientist, LLNL, USA
The original supercomputer centers were driven by the needs of scientists to simulate complex, physical phenomena. Since then, high-performance systems have been used in various data-centric applications ranging from analysis of astronomy images to web searches. The focus in supercomputing has often been on the teraflops we can obtain while running a code or the performance of single component of a system, such as file I/O. In this panel, we will look at systems, and their performance, from the viewpoint of applications. Specifically, we will look at various data-centric applications, understand the needs of these applications, identify what users would most like to see provided in their systems, and determine if current architectures and systems can meet these needs.
Questions:
For those in applications:
- Briefly describe your application(s)
- What are the two or three most frustrating architecture/systems issues that prevent you from being more productive?
For those in architecture/systems: TBD