Columbia is a 10,240-processor supercomputer consisting of 20 Altix nodes with 512 processors each, and currently ranked as one of the fastest in the world. In this paper, we briefly describe the Columbia system and its supporting infrastructure, the underlying Altix architecture, and benchmark performance on up to four nodes interconnected via the InfiniBand and NUMAlink4 communication fabrics. Additionally, three science and engineering applications from different disciplines running on multiple Columbia nodes are described and their performance results are presented. Overall, our results show promise for multi-node application scaling, allowing the ability to tackle compute-intensive scientific problems not previously solvable on available supercomputers.
Weitere Kapitel dieses Buchs durch Wischen aufrufen
- Impact of the Columbia Supercomputer on NASA Science and Engineering Applications
- Springer Berlin Heidelberg