Red Hat blog
Red Hat Enterprise Linux (RHEL) provides the operating system cornerstone for the top three supercomputers in the world according to the June 2021 TOP500 ranking.
The biannual list showcases the 500 most powerful computer systems in the world to provide a better understanding of the high-performance computing market, and encourage collaboration and the exchange of data and software throughout the tech industry, academia and research organizations.
In the top ten of the current TOP500 list, Red Hat Enterprise Linux continues to serve as the operating system for:
Fugaku, the top-ranked supercomputer in the world based at RIKEN Center for Computational Sciences in Kobe, Japan.
Summit, the number two-ranked supercomputer based at Oak Ridge National Laboratory in Oak Ridge, Tennessee.
Sierra, the third-ranked supercomputer globally, based at Lawrence Livermore National Laboratory in Livermore, California.
The top three systems in this iteration of the TOP500 run on Arm (Fugaku) and IBM Power9 CPUs and NVIDIA GPUs (Summit and Sierra). Red Hat Enterprise Linux's support of Arm, Power9 and NVIDIA GPUs enables the kind of raw performance that a supercomputing environment needs.
Fugaku boasts 7,630,848 cores, which allows it to achieve a high performance linpack (HPL) benchmark score of 442 petaflops. Summit remains the fastest supercomputing system in the United States and second fastest worldwide with a performance of 148.8 petaflops. It shares architecture with Sierra, the third ranked system on the TOP500. Both systems are assembled using standard IBM servers equipped with two Power9 central processing units (CPUs) and six NVIDIA Tesla V100 graphic processing units (GPUs). Its servers are interconnected via Mellanox dual-rail EDR InfiniBand network.
In addition to HPL, the TOP500 also incorporates high-performance conjugate gradient (HPCG) benchmark results into its report. HPCG benchmarks provide an additional metric for assessing supercomputer results and are designed to exercise computational and data access patterns that more closely match a different set of applications than HPL.
Red Hat serves as the operating system for five out the top 10 supercomputers measured by HPCG. They are:
Fugaku ranked first, achieving 16 petaflops
Summit ranked second, achieving 2.93 petaflops
Sierra ranked fourth with 1.79 petaflops
Dammam-7 ranked seventh with 0.88 petaflops
Wisteria/BDEC-01 (Odyssey) ranked ninth with 0.81 petaflops
Red Hat Enterprise Linux also powers two of the five most efficient supercomputers in the world, according to the Green500 list. Machines on the Green500 list are evaluated on both performance and the power consumed while achieving said performance. Sustainable supercomputing values finding the most energy-efficient performance possible.
In the top five of the Green500 list, Red Hat Enterprise Linux serves as the operating system for:
HiPerGator AI, the second most sustainable system in the world, is based at the University of Florida. This NVIDIA system utilizes 138,880 AMD EPYC processor cores with NVIDIA A100 GPUs and holds the 22nd spot on the TOP500 list.
MeluXina, fourth on the Green500, is located in Luxembourg. This machine is capable of running in RHEL-compatible mode, uses AMD EPYC processors and is ranked 37 on the TOP500.
We believe these rankings exhibit Red Hat Enterprise Linux’s ability to provide a reliable foundation for the most powerful computer systems available while paving the way for the next generation of supercomputing applications based on containers.
Red Hat’s technologies for exascale era
The next generation of supercomputing systems are being developed with exascale in mind. Exascale, the capability to perform a quintillion floating point operations per second, is difficult to achieve even for TOP500’s top ranked supercomputer, Fugaku, which is considered to be pre-exascale. To cross the exascale barrier and claim the title of the first exascale supercomputer in the world, a system would have to be capable of performing more than double of Fugaku’s floating point operations.
Red Hat is committed to creating better technology the open source way and has partnered with others in the supercomputing community to advance supercomputing systems towards exascale performance.
Components of Red Hat’s distributed container toolkit, including Podman, Skopeo and Buildah, are already ready for building and deploying containerized applications at exascale. Red Hat has partnered with Sandia National Labs’ ECP Supercomputing Containers Project to develop a container platform for exascale systems based on Podman and universal base image (UBI) that can run across multiple architectures, such as x86-64, Arm and POWER.
Red Hat is also involved with joint work with IBM and Lawrence Livermore National Laboratory on scheduler integration into Kubernetes, which aims to enable next-generation workloads by adding LLNL’s Flux resource and job management framework with Red Hat OpenShift. Coupling Flux with Red Hat OpenShift, along with other high-performance Kubernetes capabilities, is part of a plan to allow for successful provisioning and managing containerized workloads at exascale while also opening up a path for deploying them into public or hybrid clouds.
Red Hat is helping to create standards and commonly used interfaces for high performance and technical computing and we see our collaborations in the supercomputing community as a prime example of what enterprise open source software can help organizations achieve.
Learn more about Red Hat’s commitment to open source and collaboration and learn how you can transform your business with high-performance computing.
About the author
Yan Fisher is a Global evangelist at Red Hat where he extends his expertise in enterprise computing to emerging areas that Red Hat is exploring.