Cluster Case Studies

Performance Case Studies

Mellanox proven performance and scalability abilities are the driving forces of the most powerful compute clusters in the world, as published in the Top500 list.

National Supercomputing Center in Shanghai | Tokyo Institute of Technology - TSUBAME 2.0 | NASA AMES Research Center - Pleiades | Commissariat a l'Energie Atomique (CEA) - TERA 100 | Moscow State University Research Computing Center - Lomonosov | San Diego Supercomputing Center - Gordon | Purdue University - Carter | University of Cambridge - HPC Cloud

National Supercomputing Center in Shanghai

  • 1.27 PF sustained performance
  • Mellanox QDR InfiniBand end to end
  • 5200-nodes with Dawning TC3600 Blades and nVidia GPGPU

The Shanghai Supercomputer Center serves a number of users from all over China, and covers 20 difference fields and industries, including weather forecast, life sciences, auto design, civil engineering, etc. This Mellanox QDR cluster is #4 on the TOP500 list, and was the first Petaflop scale system in China.

(Top)

Tokyo Institute of Technology - TSUBAME 2.0

  • 1.2 PF sustained performance
  • HP ProLiant SL390s G7 1400 servers
  • 4,200 NVIDIA Tesla M2050 Fermi-based GPUs
  • Mellanox QDR InfiniBand

The Tsubame 2.0 is the first supercomputer in Japan to achieve Petascale performance. This cluster utilized large-scale GPU acceleration and massive parallel storage. All compute nodes are interconnected by low-latency, high-bandwidth full-bisection Infiniband networks, providing each node 10GB/s inter-node bandwidth.

(Top)

NASA AMES Research Center - Pleiades

  • 11,776 nodes providing 1.34 PF of sustained performance
  • Two Infiniband fabrics with all nodes connected in a partial 11D hypercube topology utilizing Mellanox HCA and switch technology

The Pleiades cluster, which ranks 7th on the TOP500 list, represents NASA's state-of-the-art technology for meeting the agency's supercomputing requirements, enabling NASA scientists and engineers to conduct modeling and simulation for NASA missions.

(Top)

Commissariat a l'Energie Atomique (CEA) -
TERA 100

  • 4,300 Bull S Series Servers
  • Mellanox QDR InfiniBand

The TERA-100, which is 10th on the TOP500, is also the fastest supercomputer and first Petascale supercomputer in Europe. CEA conducts fundamental and applied research into many areas, including the design of nuclear reactors, the manufacturing of integrated circuits, the use of radiation for curing illnesses, seismology and tsunami propagation, the safety of computerized systems, etc.

(Top)

Moscow State University Research Computing Center - Lomonosov

  • 4,400 nodes delivering 674TF sustained performance
  • T-Platforms TB2-XN T-Blade Chassis
  • Mellanox end-to-end QDR InfiniBand

Increasing price/performance, scalability, and fault tolerance were key requirements for Moscow State University in the design of their latest supercomputer. The Lomonosov supercomputer, which is #18 on the TOP500, is divided into 2 partitions with different node architecture, an x86 part with peak performance of 510TF, and a GPU part with peak performance of 863TF. In a total, Lomonosov uses six different types of compute nodes, including processors with different architectures, all interconnected with QDR InfiniBand.

(Top)

San Diego Supercomputing Center - Gordon

  • Dual-rail 3D-torus InfiniBand interconnect

The Gordon cluster, which is #48 on the TOP500, is addressing the needs of 'Big Data' in HPC. The supercomputer also utilizes a unique architecture consisting of 300TB of flash-based storage access along with a 4 PB data storage system with 100-GB/s delivery capabilities across the InfiniBand fabric.

(Top)

Purdue University - Carter

  • 648 node Intel Sandy Bridge
  • Mellanox FDR InfiniBand

The Carter supercomputer has the distinction of being the first Sandy Bridge PCIe Gen3 and FDR 56-Gb/s Infiniband cluster on the TOP500. This supercomputer achieve 187 TF sustained performance, making it the smallest node count cluster to ever achieve this mark.

(Top)

University of Cambridge - HPC Cloud

  • Dell based servers
  • Mellanox end-to-end QDR InfiniBand

The Cambridge University HPC Cloud Computer provides on-demand access to support over 400 users spread across 70 disparate research groups. These users run a variety of HPC applications on the supercomputer, including coupled atmosphere-ocean model simulations, HPC-based research such as bio-medicine, clinical-medicine and social sciences, and traditional sciences such as chemistry, physics and biology.

(Top)