Sequoia (supercomputer)

Sequoia (supercomputer)
Name	Sequoia
Manufacturer	IBM
Model	Blue Gene/Q
Installed	2011
Location	Lawrence Livermore National Laboratory
Purpose	High-performance computing
Processors	1,572,864 cores
Peak	20 petaflops (theoretical)
Architecture	IBM PowerPC A2
Os	Customized Linux

Sequoia (supercomputer) was an IBM-built Blue Gene/Q system deployed at Lawrence Livermore National Laboratory in 2011 to support computational modeling for national security and scientific research. The system combined dense IBM hardware with specialized software stacks to deliver petascale performance for applications in Los Alamos National Laboratory, Sandia National Laboratories, and collaborations with academic partners such as Stanford University, Massachusetts Institute of Technology, and University of California, Berkeley.

Overview and Design

Sequoia was designed by IBM engineers working with teams from Lawrence Livermore National Laboratory, Department of Energy, and the National Nuclear Security Administration to address simulation demands driven by the Stockpile Stewardship Program and by large-scale scientific campaigns involving Oak Ridge National Laboratory, Argonne National Laboratory, and Brookhaven National Laboratory. The project integrated lessons from the Blue Gene series, incorporating advances that had been trialed at IBM Thomas J. Watson Research Center and informed by roadmaps from the Exascale Computing Project and guidance from Cray Research and the High Performance Computing Modernization Program.

Hardware Architecture

Sequoia's hardware was a massively parallel arrangement of IBM Blue Gene/Q compute nodes leveraging the PowerPC-based PowerPC A2 processor design and a 5D torus interconnect topology similar to earlier Blue Gene/L and Blue Gene/P machines. The system comprised racks of six-node blades aggregated into midplanes, cabinets, and a full machine footprint housed at Lawrence Livermore National Laboratory facilities designed with cooling and power infrastructure influenced by standards from the U.S. Department of Energy and building practices at Sandia National Laboratories campuses. Each compute card and node employed components sourced from suppliers including GlobalFoundries, Samsung Electronics, Micron Technology, and Intel-affiliated fabrication partners, while network control leveraged technologies evolved from InfiniBand research and switch designs documented in publications from IEEE and ACM conferences.

Performance and Benchmarks

Sequoia achieved a peak theoretical performance of about 20 petaflops and sustained LINPACK results that placed it atop the TOP500 list in 2012, joining other leading systems like Jaguar (supercomputer), Tianhe-1A, and later K computer and Titan (supercomputer). Benchmarks conducted by teams from Lawrence Livermore National Laboratory and validated by independent reviewers from Oak Ridge National Laboratory and Argonne National Laboratory measured performance on scientific kernels derived from codes developed at Los Alamos National Laboratory, Sandia National Laboratories, and academic groups at University of Illinois Urbana–Champaign, University of Texas at Austin, and Georgia Institute of Technology. LINPACK metrics, as well as application-level benchmarking using suites from NAS Parallel Benchmarks and tests influenced by SPEC and HPCC, highlighted strengths and bottlenecks tied to memory bandwidth, network latency, and I/O subsystems designed around parallel filesystems like Lustre.

Software and Programming Environment

The Sequoia software ecosystem combined a customized Linux distribution maintained by IBM and Lawrence Livermore National Laboratory engineers with compilers and libraries from IBM XL, GNU Compiler Collection, and runtime systems developed in collaboration with research groups at University of California, Davis and University of Michigan. Programming models supported included MPI implementations refined by contributors from Argonne National Laboratory and Los Alamos National Laboratory, threading with OpenMP as used in projects from Sandia National Laboratories, and task-parallel frameworks inspired by research at Carnegie Mellon University and University of Illinois Urbana–Champaign. I/O and data management relied on middleware and storage strategies developed with partners at National Energy Research Scientific Computing Center and used tools like HDF5 and NetCDF common in projects at NOAA and NASA centers including NASA Ames Research Center.

Deployment and Operational History

Deployed in 2011 and brought to full production in 2012, Sequoia entered operational service under management structures coordinated by Lawrence Livermore National Laboratory, the National Nuclear Security Administration, and oversight from the U.S. Department of Energy Office of Science. The machine supported classified and unclassified workloads for agencies including Department of Defense laboratories and scientific initiatives involving collaborators at Harvard University, Yale University, Princeton University, Columbia University, and the University of California system. Operations teams adopted maintenance practices influenced by procedures at Oak Ridge National Laboratory and configuration management approaches from Sandia National Laboratories. Over its service life, Sequoia was subject to upgrade cycles, security reviews aligned with National Institute of Standards and Technology guidance, and eventual decommissioning planning coordinated with peer centers such as National Center for Supercomputing Applications.

Applications and Scientific Impact

Sequoia enabled large-scale simulations across domains championed by researchers at Lawrence Livermore National Laboratory, including computational chemistry projects connecting to work at California Institute of Technology, climate modeling collaborations with NOAA Geophysical Fluid Dynamics Laboratory, and astrophysics simulations tied to teams at Princeton Plasma Physics Laboratory and Fermi National Accelerator Laboratory. It powered codebases such as ALE3D and other multiphysics packages developed with contributions from Los Alamos National Laboratory and Sandia National Laboratories, and facilitated research outputs published by scientists affiliated with American Physical Society and conferences hosted by IEEE Computer Society and ACM SIGARCH. By enabling ensemble simulations and high-resolution models, Sequoia influenced efforts in materials science at Argonne National Laboratory, fusion energy studies at Princeton Plasma Physics Laboratory, and nuclear nonproliferation analysis coordinated with National Nuclear Security Administration programs. The system also served as a platform for software engineering research at institutions like University of Washington, University of California, San Diego, and University of Southern California.

Category:Supercomputers