AEI, MEGWARE and ASUS deliver extreme computing power for analysis of gravitational-wave data

Overview The Max Planck Institute for Gravitational Physics, also known as the Albert Einstein Institute (AEI), is the world’s largest research institute in general relativity and beyond. The institute is located in Potsdam-Golm and in Hannover, where it is closely related to the Leibniz Universität Hannover. Gravitational waves are a consequence of Albert Einstein's general theory of relativity – tiny ripples in space-time, generated in stellar explosions, the mergers of compact stellar remnants and other cataclysmic cosmic events. Scientists have built a worldwide network of large-scale detectors to receive these waves. In order to analyze the massive-detector data streams, high-performance computing clusters are required. That’s why the AEI is the home of the Atlas Computing Cluster, the world's largest and most powerful computing resource dedicated to gravitational-wave data analysis.

The Project

At the initiation of the Atlas project, MEGWARE and server manufacturer ASUS provided barebone-server samples to AEI to enable the institute to evaluate the quality, flexibility and power of ASUS equipment. Previous experience of Debian GNU/Linux prompted AEI to deploy this distribution across the cluster’s servers.

The combined efforts of AEI, MEGWARE and ASUS proved successful and the Atlas Computing Cluster is now at the forefront of the HPC world, proving that ASUS is able to meet the needs both of the largest customers and the most specialized applications. As Atlas cluster administrators Dr. Carsten Aulbert and Dr. Henning Fehrmann stated: “Due to the very good cooperation with MEGWARE and ASUS, AEI's Atlas Computing Cluster is now by far the largest dedicated resource for gravitational-wave analysis.”

Specifically, the Atlas Computing Cluster contains a total of 2,365 ASUS RS100-E8/PI2 servers — 2,300 powered by with Intel® Xeon® E3-1220v3 quad-core processors and 65 driven by Intel Xeon E5-1650v2 six-core processors.

A subset of 460 of these nodes are equipped with high-end 2GB GeForce® graphics cards. These GPU-enhanced nodes benefit from the latest high-density GPU co-processors and deliver high-density computing power, scalable expansion capabilities, high speed, great interconnectivity and intelligent thermal designs. With ASUS servers providing both the small compute nodes and GPU nodes for extreme computing performance, the Atlas Computing Cluster delivers 500 teraflops/s CPU peak performance and 350 teraflops/s GPU peak performance.

Clearly, the flexibility and power of ASUS servers allows researchers to run a range of applications with optimized performance and make the most of leading HPC-acceleration technologies. ASUS and MEGWARE also delivers high-quality after-sales service, with fast response times — protecting that our customers’ investments in ASUS equipment servers and ensuring steadfast operation. The predecessor to the Atlas system was developed and installed until 2008. However, the capabilities of the original were soon exhausted and the opportunity of a joint tender with ASUS proved compelling to MEGWARE. The result was a success for both company and region: "This supercomputer, manufactured in Chemnitz, Saxony, demonstrated first-class energy efficiency, and an extremely high throughput," said André Singer, Project Manager at MEGWARE. Jörg Heydemüller from MEGWARE´s sales team added: "The fact that this major supercomputer contract has been awarded to Saxony is a great honor and decisive for the high-performance computing community.”. “It’s also great news for major events, such as the HPC User Meeting in Chemnitz and Leipzig’s International Supercomputing Conference.”