On XSEDE, what is Keeneland?
On the Extreme Science and Engineering Discovery Environment (XSEDE), Keeneland is a hybrid CPU/GPGPU system intended for codes that can take advantage of GPU accelerators. Keeneland is hosted by the Georgia Institute of Technology (Georgia Tech), and administered by the National Institute for Computational Sciences (NICS).
The Keeneland Initial Delivery System (KIDS), presently available as a production XSEDE digital service, is composed of an HP SL-390 (Ariston) cluster, with Intel Westmere hex-core CPUs, NVIDIA 6 GB Fermi GPUs, a QLogic QDR InfiniBand interconnect, and two hex-core CPUs and three GPUs per node, for a total of 120 nodes, 240 CPUs, and 360 GPUs. For detailed specifications and user information, see Georgia Tech Keeneland User Guide.
In September 2012, KIDS will be replaced with the Keeneland Final System (KFS), which will include:
- 264 nodes (KIDS has 120 nodes)
- A Mellanox FDR Interconnect (KIDS has a Qlogic QDR Interconnect)
- 16 CPU cores per node (KIDS has 12 CPU cores per node)
- Intel Sandy Bridge CPUs, which support PCI Express 3 (KIDS has Intel Westmere CPUs)
The Keeneland project is a five-year, $12-million National Science Foundation (NSF) grant awarded to Georgia Tech and its partners, the University of Tennessee at Knoxville (UT), the Oak Ridge National Laboratory (ORNL), and NICS, for the deployment of an innovative high-performance computing system, bringing emerging architectures to the open science community. For more, see Georgia Tech's Keeneland overview.
Note: Support for XSEDE users of Keeneland is coordinated through NICS via the XSEDE Help Desk.
For more about XSEDE compute, advanced visualization, storage, and special purpose systems, see the Resources Overview, Systems Monitor, and User Guides. For scheduled maintenance windows, outages, and other announcements related to XSEDE digital services, see User News.
This document was developed with support from National Science Foundation (NSF) grant OCI-1053575. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the NSF.
Last modified on January 03, 2013.