|Name:||The Sequoia System & Facilities Integration Story|
|Time:||Thursday, June 21, 2012
9:30 AM - 9:45 AM
CCH - Congress Center Hamburg
|Speakers:||Kim Cupps, LLNL|
|Abstract:||Sequoia, a peak 20PF/s Blue Gene/Q system, will serve National Nuclear Security Administration's (NNSA) Advanced Simulation and Computing (ASC) program. Computational resources are essential to fulfill stockpile stewardship requirements through simulation science.
Problems at the highest end of this computational spectrum are a principal driver for the ASC Program as highly predictive codes are developed. Sequoia is an Uncertainty Quantification (UQ)-focused computing system currently being integrated at Lawrence Livermore National Laboratory. Sequoia will simultaneously run integrated design code calculations and science materials calculations enabling sustained performance of 24 times ASC’s previous Purple calculations and 20 times ASC’s Blue Gene/L calculations.
LLNL has been preparing for delivery of Sequoia for well over three years. During the past nine months we have been consumed with the integration challenges of siting the facilities and infrastructure for Sequoia, as well as the machine itself. Integration of Sequoia continues, acceptance testing will begin in early July and production level computing is expected in January of 2013. This talk focuses on the facilities and system integration victories and challenges of siting Sequoia. Application analysis on Blue Gene/Q simulators and early hardware has been underway to tackle the issues associated with running on this 1.5 million core machine.