SHOW PREVIEW: SC10
➤
GPU appliance capable of housing up to eight Tesla GPUs supporting up to four server nodes. NextIO will demonstrate ‘drag and drop’ hot plug reassignment of GPUs to a server without bringing down GPU applications on that server. The server does not have to be rebooted, the GPU applications do not stop and the server enclosure does not have to be opened in order to add or remove GPU resources to a server node. The appliance provides a reconfigurable pool of GPU resources.
www.nextio.com
Some of the exhibits and
demonstrations at NICT Japan’s booth include:
joint research with Kyushu and Ehime Universities, demonstrating real-time space weather forecast calculation, data transfer, visualisation with GPU clusters, and the JGN2plus network; a joint research project with GLIF, AIST, and KDDI R&D laboratories, showing dynamic multipoint circuit provisioning over Japan, Korea, and US multi-domains; collaboration between dynamic circuit provisioning and applications such as TDW; and joint research with NTT Laboratory, which will be showing application-initiated network path provisioning and real-time measurement and visualisation of very precise streaming characteristics. Finally, the JGN2plus network, Japan’s largest test-bed network, and its research topics will be introduced.
www.nict.go.jp
Numascale will present and demonstrate the newly-launched NumaConnect SMP adapters.
Based on the NumaChip, NumaConnect enables the HPC community to build highly scalable shared memory machines (SMP) in hardware at a fraction of the cost of current alternatives.
NumaChip is a combined cache coherence controller and distributed switching fabric interfacing to AMD’s coherent HyperTransport. NumaChip uses a directory-based cache coherence protocol that is far more scalable than broadcast
40
or snooping based protocols. The on-chip switch connects to neighbour nodes in 2- or 3D torus topologies and can scale to 4,096 nodes. The remote caches can be configured up to 16Gb per node and can support up to 256TB of RAM. With NumaConnect and commodity servers, Numascale makes scalable SMP systems available at cluster pricing to the benefit of research, academia and industry users.
www.numascale.com
The Nvidia Tesla 20-series is designed from the ground up for high-performance computing. Based on the next generation Cuda GPU architecture codenamed ‘Fermi’, it supports many ‘must have’ features for technical and enterprise computing. These include ECC memory for uncompromised accuracy and scalability, support for C++, and 8x the double precision performance compared to Tesla 10-series GPU computing products. When compared to the latest quad-core CPU, Tesla 20-series GPU computing processors deliver equivalent
together with DEISA, the Distributed European Infrastructure for Supercomputing Applications, takes place on Wednesday 17 November from 12.15 to 1:30pm. The session will present the latest news about the PRACE RI. It covers the current status and the future plans; the results to be expected during the EC-funded implementation projects, the integration of services currently provided by DEISA within the European HPC ecosystem, and collaboration opportunities for academia and industry.
www.prace-ri.eu
SGI will be featuring a number of storage solutions for both scale-
performance at 1/20th the power consumption and 1/10th the cost. Each Tesla GPU features hundreds of parallel Cuda cores and is based on the Nvidia Cuda parallel computing architecture with a rich set of developer tools (compilers, profilers, debuggers) for popular programming languages APIs like C, C++, Fortran, and driver APIs like OpenCL and DirectCompute.
www.nvidia.com
PRACE, the Partnership for Advanced Computing in Europe, is building a persistent pan-European Research Infrastructure (RI) for providing leading HPC services to enable world-class science and engineering for European academia and industry. The first production system, a 1-petaflop/s IBM BlueGene/P (Jugene) at FZJ (Forschungszentrum Jülich) is available for European scientists. The PRACE BoF at SC10 ‘PRACE – The European HPC infrastructure created’
SCIENTIFIC COMPUTING WORLD OCTOBER/NOVEMBER 2010
up and scale-out HPC environments, along with products for persistent data management. Key among these are the SGI InfiniteStorage 16000 and the Copan 400. The InfiniteStorage 16000 is a next generation RAID system, which meets the needs of both bandwidth-hungry and IOPs-heavy HPC workloads. Combined with server platforms like Altix UV and the CXFS shared file system, SGI can deliver I/O throughput well above and beyond typical competitive end-to-end offerings. The Copan 400 is SGI’s MAID (Massive Array of Idle Disks) platform, which delivers cost-effective storage of long term data similar to tape systems, but with the reliability and performance of disk. SGI will also be showing its latest server offering, which has an architecture specifically for optimising hundreds to thousands of high-bandwidth PCIe slots. Codenamed ‘Project Mojo’ the platform will house this year’s and future, PCIe-based accelerators, for maximising the amount of compute possible in a single cabinet.
www.sgi.com
Whamcloud will be exhibiting at the show. The company is formed from worldwide high-performance computing (HPC) storage industry veterans. It is focused on enabling application scaling and information insight through the evolution of HPC storage technologies in collaboration computing centres.
www.whamcloud.com
www.scientific-computing.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52 |
Page 53 |
Page 54 |
Page 55 |
Page 56