Looking at bare numbers, SUSE might be known as the # 2 Linux distribution worldwide. But in many fundamentally important market segments, we provide the leading Linux operating system. Among these areas are mainframe computing, SAP on Linux, Linux in Retail, and: High Performance Computing.
You might not always see us in the front raw here, or being in the spotlight anywhere anytime – well, you could argue we are “just the operating system” 😁. But, the operating system is crucial for an HPC system, it needs to provide excellent performance. And we ARE there, working behind the scenes. For us it is quite important that the history of Linux in HPC in general and the history of SUSE in HPC really go hand in hand. SUSE was a very early player in the Linux space (to be precise: we are the oldest Linux distro in the market, our engagement dates back to 1992, and this year we can celebrate our 25th anniversary!). We’ve also always provided Linux code to the HPC market, and SUSE engineers have made significant contributions to the advancement and tuning of the Linux kernel and key kernel-related performance technologies.
Now, just make an educated guess: how many of the most important supercomputers run an operating system version of SUSE? Not even a month ago, my colleague Jeff Reser in his blog “ISC, SUSE, and the Top500” highlighted the latest release of the TOP500 supercomputer list, published during the International Supercomputing Conference in June in Frankfurt, Germany. This list ranks the world’s most powerful supercomputers – and he states that half of the Top50 of them is powered by SUSE’s Linux distributions!
Of course we are also very thankful and proud of our close cooperation and good relationships with other players in the HPC world, including technology partners and supercomputing sites. As an example, Jeff in the just mentioned article showcased MareNostrum at the Barcelona Supercomputing Center in Spain, one of my favorite supercomputers. In previous blog posts, I highlighted Pleiades from NASA in the US, or Piz Daint in Switzerland as the fastest supercomputer in Europe. And we have done success stories about famous supercomputers such as SuperMuc at Leibniz Supercomputing Centre in Germany, Pangea from Total in France or Tsubame at the Tokyo Institute of Technology.
But so far we never looked Down Under. This changes today – and it’s definitely time! Good HPC news reached us from Australia: the Commonwealth Scientific and Industrial Research Organization (CSIRO) has recently deployed and put into production its newest supercomputer. It is named Bracewell – after the very famous Australian Physicist, Mathematician and Radio astronomer Ronald Bracewell, who worked for CSIRO in the early 50ties.
Some technical details (just skip reading this paragraph if this is going too deep): Bracewell is based on the SUSE Technology Partner Dell EMC PowerEdge platform. It boasts 114 PowerEdge C4130 servers, hooked together with 100Gbps EDR InfiniBand interconnects. Each server is equipped with four NVIDA P100 GPUs and two Intel Xeon 14-core CPUs. All in all, it contains 1,634,304 Cuda compute cores, 3192 Xeon compute cores, and 29TB of RAM. The GPUs alone represent over 2.4 petaflops of peak performance. From a flops perspective, that would easily make it the most powerful supercomputer in Australia. On the most recent top500 list, it would very likely be placed around rank #50, while the current most powerful Australian supercomputer Raijin ranks #70.
Bracewell mostly runs Linux, but nodes can be dynamically provisioned to run Windows when required by researchers (… dual-boot machines are a quite good option in HPC …). What Linux, you ask? No question, it is powered by SUSE Linux Enterprise Server 12, managed from Bright Computing’s Bright Cluster Manager – another valued SUSE partner. With the latest service pack update, CSIRO has even seen a 10% improvement in I/O. (BTW, its “little brother” Pearcey – a CPU only cluster – is also running SUSE Linux Enterprise Server. Pearcey just had an expansion, too, of a bunch more nodes.)
One of the first huge computing tasks for Bracewell is boosting the bionic vision research that aims to restore sight for people with profound vision loss, through new computer vision processing that uses large scale image data sets to optimize and learn more effective processing.
In addition, the new supercomputer will provide computational support for science projects around medical diagnosis and therapeutic treatments (not too long ago, I wrote a blog about this highly important research area for HPC), traffic and logistics optimization, modeling of new material structures and compositions, and machine learning for image recognition and pattern analysis.
Machine learning and Artificial Intelligence are hottest topics in HPC. Let me – once again – point to my colleague (and HPC Product Marketing Manager) Jeff Reser, and cite from his blog “AI and HPC – A Match Made For Technological Singularity?” where he states:
“The adoption of high performance data analytics, big data and a more cohesive open source HPC stack has fundamentally changed the landscape for building AI and deep machine learning applications. Along with a strong HPC operating system and infrastructure coupled with advancements in AI, machines are becoming more intelligent at a faster pace. They are improving their ability to recognize patterns, iteratively learn from data analysis and sharpen predictive analytics. Industries starting to benefit from these AI applications include banking and finance, pharmaceuticals, shipping and transportation, healthcare, manufacturing and agriculture.”
Bracewell – welcome to the family of supercomputers powered by SUSE Linux Enterprise Server, working on solving the most challenging problems for a better life and a future worth living. We will keep an eye on you 😉.
Disclaimer: The text at hand has not been reviewed by a native speaker. If you find typos or language mistakes, please send them to me (firstname.lastname@example.org) – or if you like them, just keep them and feed them. 😆