Keeneland: Computational Science Using Heterogeneous GPU Computing
Jeffrey S. Vetter, Richard Glassbrook, Karsten Schwan, Sudhakar Yalamanchili, Mitch Horton, Ada Gavrilovska, Magda Slawinska, Jack J. Dongarra, Jeremy Meredith, Philip C. Roth, Kyle L. Spafford, Stanimire Z. Tomov, John Wynkoop · 2017
The Keeneland Project is a five-year Track 2D grant awarded by the National Science Foundation (NSF) under solicitation NSF 08–573 in August 2009 for the development and deployment of an innovative high performance computing system. The Keeneland team originally assessed numerous technologies to propose in 2008. Keeneland Initial Delivery System has been installed, and operating since November 2010; it is primarily used for the development of software tools and preparation of applications to use this innovative architecture. The Keeneland Full Scale System has been installed and operating since October 2012; it is an XSEDE production resource, which is allocated quarterly by XSEDE&s;s XRAC allocations committee. The system software used on the Keeneland systems reflects the perspective that the systems are Linux x86 clusters with GPUs as compute accelerators. The Keeneland programming environment is a blend of standard HPC software, such as MPI, augmented by GPU-enabled programming environments.