International Supercomputing 2009

The International Supercomputing Conference (ISC) is Europe’s leading conference and exhibition on high performance computing, networking and storage. 1,500 attendees and 120 exhibitors are expected over the four days

Blade Network Technologies

Datacentres are running out of power. With this at the forefront of industry thinking, at ISC'09 Blade Network Technologies (www.bladenetwork.net) will demonstrate performance versus power consumption of various interconnect options and will provide insight on switching fabric optimisation and consolidation.

Blade's mature operating system provides capabilities to efficiently consolidate storage and data networking fabric over 10GbE. Blade will also be unveiling its achieved performance benchmarks using its ultra-low latency 10 Gigabit Ethernet (10GbE) switch portfolio with leading HPC applications, such as: PAM Crash, Fluent, VASP, RADIOSS and many more. Its 10GbE portfolio meets the requirements for highly demanding computing environments and offers the best price performance solutions.

Blade will showcase a range of industry first products in its Cloud Ready Network infrastructure. These latest products provide a truly open and flexible datacentre environment, allowing network resources to be deployed and moved in the simplest way possible.

ClusterVision

ClusterVision will be exhibiting and demonstrating: the ClusterVisionOS cluster management software, the SiCortex energy-efficient supercomputer, and the Cray CX1 deskside supercomputer.

ClusterVisionOS is a Linux-based cluster operating system and management environment which greatly simplifies the management and use of HPC clusters. With ClusterVisionOS, a cluster administrator can install, manage and use multiple clusters simultaneously without the need for expert knowledge of Linux or HPC.

ClusterVisionOS provides the system administrator with a single system view while offering the user an excellent environment for HPC. Based on open standards, ClusterVisionOS combines the best of open source and commercial software for HPC. It has been designed to scale to thousands of nodes and to interoperate with common grid technologies and application APIs. Furthermore, it has been Intel Cluster Ready certified, which ensures compatibility with a wide range of certified applications.

Computational Science and Engineering Department, Science and Technology Facilities Council

The Computational Science and Engineering (CSE) Department of the Science and Technology Facilities Council (www.cse.scitech.ac.uk) acts as a UK focus for the development, application and support of research in computational science and engineering.

The institution has more than 50Tflops of compute power on-site, comprising two IBM Blue Gene systems, a GPU system and HPCx, one of the UK's national supercomputers, plus several other systems. The site is a partner with the Edinburgh Parallel Computing Centre (EPCC) in the running of both HPCx and HECToR, the UK's latest resource for academic computational science and engineering research.

CSE's research includes advanced research computing, band theory, computational engineering, computational materials, distributed computing, mathematical algorithms and sub-routines, molecular simulation and quantum chemistry, plus scaling codes to operate on tens of thousands of processors.

At ISC, CSE will overview its work with UK academics, focusing on scientific highlights from the collaborative computational projects and its high-performance computing activities.

Convey Computer Corporation

Convey Computer Corporation (www.conveycomputer.com) will demonstrate a hybrid-core computer, the Convey HC-1. Technology highlights include up to 92 per cent lower power and cooling costs, simplicity of programming with ANSI-standard C, C++, and Fortran, and reloadable personalities (instruction sets) that increase productivity for CPU-intensive applications.

Convey's system, described as 'a remarkably ingenious and innovative HPC architecture', combines general-purpose multi-core processors with special-purpose FPGAs and leverages the x86/Linux ecosystem.

Hybrid-core computing saves on power costs, increases compute performance for certain applications, and enables customised applications to be developed in months – not years.

Dell

ISC09 brings together users of HPC from research and industry. The common ground is an increasing demand to drive outputs and innovation. The need to simplify HPC has never been greater as we seek to solve global problems by scientific means and address current business demands within the competitive landscape of a challenging world economy. To support customers and meet these demands, Dell (www.dell.de) will show the ideal HPC building block - a new rack server model for HPC in an Intel Cluster Ready configuration. Dell will be demonstrating how virtualisation within its latest Blade Server system delivers a Virtual SMP making traditionally high-end only supercomputing features available to many more business and scientific innovators. The company will show how GP/GPU technology provides the opportunity for improving output within a standards-based HPC. Dell will be talking with customers about how the company simplifies HPC while driving the cutting edge of business and science.

DKRZ

The German Cloud Computing Centre (DKRZ) (www.dkrz.de) is a national German service centre for earth system research. DKRZ provides: high performance computing, data services and a climate database, visualisation of model-results and domain specific qualified user support. Climate simulations depend on access to high performance computers.

DKRZ is a worldwide unique high performance computing centre providing dedicated resources and services for earth system modelling. It operates a new IBM-Power6 system, which delivers more than 140 Teraflop/s in peak performance.

DKRZ computers are used for complex model calculations whose results contribute, for example, to the Assessment Reports of the Intergovernmental Panel on Climate Change. At ISC'09 DKRZ will present the results of the climate change simulations on an Omniglobe - an 'interactive planet' with an unique 360-degree view.

DKRZ also informs about its newly-established research in the fields of energy efficiency in high performance computing.

Finisar

Finisar (www.finisar.com) is a technology leader for fibre optic subsystems and network test systems enabling high-speed voice, video and data communications for networking, storage, wireless, and CATV applications. For more than 20 years, Finisar has provided critical optics technologies to system manufacturers to meet the increasing demands for network bandwidth and storage. In 2007, Finisar launched its Active Optical Cable family including Laserwire for 10GbE and Quadwire for 40GbE and InfiniBand QDR.

Fujitsu

Fujitsu (www.fujitsu.com) integrated Fujitsu Siemens Computers in April 2009 and established Fujitsu Technical Solutions. Fujitsu will continue to be engaged in the client and enterprise platform business in EMEA through FTS by delivering high value-added technology solutions. Fujitsu has been leading the HPC market fore more than 30 years and is currently developing multi petaflops class supercomputer with unique processor, interconnect and packaging technologies. Fujitsu is also contributing to Japan's next generation supercomputer project led by RIKEN with these cutting-edge technologies. A broad range of computing products, such as x86-based cluster, SPARC-based SMP server, software and solutions, meet comprehensive technical computing requirements. SPARC64 VIIIfx, Fujitsu's new high-end CPU, the Primergy BX900 Blade system, a single 10U chassis accommodated up to 18 server blades and the SynfiniWay middleware for building clouds will be highlighted at ISC09.

Gridcore AB

Gridcore AB (www.gridcore.se) is a privately-held company based in Gothenburg (Sweden) that owns and operates Gompute, a high performance computing on demand service for technical and scientific users.

Gompute's customers ranges from one man company to multinational corporations operating in sectors as energy, industry, academia and so on. Gompute operates worldwide with clusters installed both in the USA and Sweden and has become a leader brand in terms of HPC on demand as easily provides users with pre-post processing and computer power.

Gompute supports the most of the CAE applications including Ansys and first-class support for OpenFOAM. Gompute's technology and services provide also with portable supercomputing where users may deploy their own hardware cloning locally Gompute environment.

IBM

IBM (www.ibm.com/deepcomputing) will demonstrate its current HPC solutions built on leading server, storage, and networking technologies. IBM will show solutions around IBM BlueGene, IBM Power 575 servers with POWER6 processors, System x iDataplex, HPC File System Management solutions and blade servers based on x86 and POWER architectures.

Furthermore IBM experts will deliver presentations on issues facing the HPC community, and participate in discussion roundtables: On Thursday 25 June, Professor Gunter Dueck, IBM Distuinguished Engineer, will deliver a keynote on 'Lean Brain Management – More Success and Efficiency by Saving Inteligence'.

At a discussion roundtable on Wednesday 24 June, David Turek, vice president Deep Computing, will explain IBM's view on the question 'Cloud Computing & HPC – Synergy or Competition?'. He will also talk about 'The Journey to Exascale Computing' at the Exhibitor Forum: an exascale computer is widely believed to be an achievable goal around the end of the next decade. This talk will briefly touch on the technological impediments, potential solutions, and ultimate payoff associated with this ambition. Dr Don Grice, IBM Distinguished Engineer, will focus on the topic 'Heterogeneous Computing – Coping with the Energy Wall' at a Hot Seat Session.

LSI

LSI (www.lsi.com/hpc) will showcase its Engenio DE6900 high-density SATA drive enclousure. Based on its sixth-generation architecture, the DE6900 is a 4U, 19-inch rack-mountable drive enclosure that enhances the ability of its 7900 HPC system to handle the massive data requirements of HPC applications. Visitors will be able to experience the serviceability, increased density, throughput and productivity offered by the enclosure which delivers a scaling capacity of 480 SATA drives within the same footprint as the current 16 disk fibre channel enclosure.

HPC storage and file systems are required to provide performance and reliability to deliver the ground-breaking work and research demanded by HPC applications and LSI will also demonstrate a high-performance file system showing how the high density 7900 increases productivity with less infrastructure.

Mellanox Technologies

Mellanox Technologies (www.mellanox.com) will showcase its complete 40Gb/s end-to-end connectivity portfolio for data centres and high-performance computing systems.

Mellanox ConnectX 40Gb/s InfiniBand adapters enable the fastest transaction latency, as low as 1usec, and the highest transaction rate of above 40 million messages per second, making it the most scalable and suitable solution for current and future transaction-demanding applications. ConnectX also delivers leading low-latency 10 Gigabit Ethernet capabilities, providing end users with ultimate protocol flexibility.

Mellanox's complete line of 36, 324, and 648-port 40Gb/s InfiniBand switches incorporate advanced tools that simplify networking management and installation, and provide the needed capabilities for the highest scalability and future growth.

Mellanox BridgeX Gateways provide the highest levels of integration and flexibility with support for 40Gb/s InfiniBand, 10 Gigabit Ethernet and up to 8Gb/s Fibre Channel interfaces, enabling seamless 40Gb/s access to existing LAN and SAN infrastructures.

Mercury Visualization Sciences Group

Mercury Visualization Sciences Group (VSG) (www.vsg3d.com) has just launched Avizo Wind, a new software package for simulation post-processing. Avizo Wind is a new edition of the Avizo 6 product line for 3D visualisation and analysis for scientists and engineers addressing geosciences, CAE, materials science, climate simulation, visualisation centers and VR facilities.

Mercury VSG is also releasing the brand new version 8 of its well-known 3D graphics toolkit Open Inventor for high-performance 3D applications development. Open Inventor 8 by Mercury is also introducing a new release of the VolumeViz, ScaleViz and MeshViz extensions.

These major releases deliver even higher performance in large data management, extended support for GPGPU, Cuda and generic computing interoperability. High-performance 3D visualisation tools using multi-core and GPGPU technologies accelerate analysis workflows, as illustrated by CFD simulation field extraction, seismic volume visual processing, and climate simulation post-processing.

Nema Labs

Nema Labs (www.nemalabs.com) offers technology to migrate software to multi-core platforms with minimal effort. It does so by offering programming tools that significantly reduce the effort needed to reliably convert sequential C/C++ programs to parallel versions. Indeed, using Nema Labs tools, programmers can be shielded from the mystery of parallelisation – through intuitive feedback from the tools, programmers modify the sequential code in a controlled, step-by-step fashion and end up with code that can be automatically parallelised and that works correctly after being validated by the tools. The tools are built on cutting-edge technology that has sprung from academic research led by Professor Per Stenstrom of Chalmers University of Technology – a pioneer in the multiprocessor architecture community. Nema Labs first product line is FASThread, which integrates into popular IDEs.

Panasas

Panasas (www.panasas.com) will be introducing the Panasas ActiveStor Family, a comprehensive and scalable suite of ultra-high-performance storage products. The Panasas ActiveStor 7, 8 and 9 Series Performance Module building blocks provides three options to meet a wide range of application requirements. The industry-leading 9-Series features new solid state technology combined with rotational disks that provide exceptional performance in terms of IOPS and bandwidth throughput in a single virtual storage image. Panasas ActiveStor systems are comprised of individual performance blades, which aggregate performance in a linearly-scalable system. This architecture provides distributed and balanced data path I/O eliminating performance bottlenecks. The Panasas ActiveScale File System provides a scalable architecture capable of achieving performance in the hundreds of GB/sec by simply adding as many individual performance modules, or racks, to achieve aggregate performance. All aggregated performance modules are presented as a single system featuring a global name space providing unified management.

Platform Computing

Platform Computing (www.platform.com) will unveil its new Platform HPC Workgroup Manager product, an advanced and complete cluster management solution. It includes all the cluster management tools needed to easily deploy, run and manage an HPC environment. The result is a complete cluster solution that provides superior performance and reliability while removing the complexity associated with Linux-based clusters. In addition, Platform is currently developing solutions to enable organisations to provide a cloud computing service within a dynamic HPC environment to internal users. At ISC, Platform will be discussing the capabilities and future of cloud computing and how its solutions provide HPC resources on-demand to increase compute capacity while reducing HPC operation costs.

Prace

Prace (www.prace-project.eu), the Partnership for Advanced Computing in Europe, will present the Prace Award 2009 at the ISC’09 opening session. Prof Dr Achim Bachem, Prace project coordinator and chairman of the board of directors, Research Center Jülich, will present the award, which goes to a team from the Supercomputing Center of Galicia for their paper on 'High Scalability Multipole Method. Solving Half Billion of Unknowns'.

The organisation will also hosts a networking session on Wednesday 24 June, entitled 'Prace: HPC for scientific breakthroughs'. Dr Thomas Eickermann, Prace project manager, FZJ, will give an update on the status of the project. Prof Dr Thomas Lippert, FZJ, will show how simulations on supercomputers led to one of the top 10 scientific breakthroughs in 2008. Dr Sergi Girona, Barcelona Supercomputing Center, talks about novel architectures in HPC, especially the exploitation of Cell processors. Dr Stefan Wesner, HLRS, motivates hybrid computing systems driven by a concrete industrial application.

Prace will also have its own booth.

Promise Technology

Promise Technology (www.promise.com) will display a full range of entry-level and enterprise-level fully redundant RAID storage systems. The storage systems use high-performance SAS hard drives as well as high capacity SATA disk drives, offering flexibility when it comes to meeting customer and application requirements. The demand from HPC solutions for a high I/O rate, high data transfer in a stable and reliable IT environment is met by using multi-channel SAS x4 storage system from the VTrak product range connected to high performance and reliable NEC servers. NEC's LXFS parallel file system speed and I/Os meet the demand for high performance applications.

SGI

SGI (www.rackable.com), the global expert in large-scale clustered computing, clustered storage, HPC and data centre enablement and services, will be launching a number of products as well as demonstrating its new Molecule prototype. SGI's CTO Dr Eng Lim Goh will be answering questions from industry experts in the International Supercomputing Hot Seat session as well as providing an executive briefing.

Shanghai Supercomputer Centre (SSC)

Shanghai Supercomputer Centre (SSC) (www.ssc.net.cn) was founded in December 2000. This facility was funded and constructed by the Shanghai Municipal Government. With first-rate facilities and full-functions, Shanghai Supercomputer Center is the first high performance computing platform open to the public.

A major breakthrough this year is that in June 2009, it will have a major new supercomputer installed, featuring more than 200 Tflops of power, placing it among the top supercomputers in the world.

Spectra Logic

Spectra Logic (www.spectralogic.com) will unveil the Spectra T950 library, which provides superior features and PB of storage with a 10,050 tape maximum capacity. With the Spectra T950 HPC users remain in constant control of their archival processes, easily and cost effectively.

HPC environments continually test the limits of technology, requiring peak performance from computing and storage equipment. The Spectra Logic T950 library supports HPC users as they push the boundaries of research and exploration, giving cost-effective storage that meets the strictest performance, growth and environmental needs.

The Spectra T950's small 'footprint' occupies approximately 61 per cent less data centre floor space than competing storage libraries. The T950 uses the least power per terabyte of data stored - translating into a 50 to 84 per cent energy usage.

The T950's open architecture allows applications to directly interface with tape storage, minimises downtime with component redundancy and advanced library management features, and offers remote management though a secure, web interface.

The Portland Group

The Portland Group (www.pgroup.com/accelerate) will demonstrate PGI Release 9.0, the first production release of the PGI Fortran and C compilers to support the PGI Accelerator Programming Model. The Accelerator Programming Model uses compiler directives to specify data and regions of code in Fortran and C programs that can be offloaded from a host CPU to an attached accelerator. The model is portable across operating systems and various types of host CPUs and accelerators. The directives allow a programmer to migrate applications incrementally to accelerator targets using ISO/ANSI standards-compliant Fortran or C. This initial release includes support for Nvidia GPU accelerators running under 64-bit Linux on x64 hosts. In addition, PGI will be discussing the first release of the Cuda Fortran language specification developed in cooperation with Nvidia.

TotalView Technologies

TotalView Technologies (www.totalviewtech.com) will be sponsoring a BOF entitled 'Multiple Ways to Debug: Users' Tips & Tricks'. This is an opportunity for users to learn about the many ways that other users are able to debug using the TotalView debugger. Users are welcome to present any TotalView tips and tricks that they have discovered. They are particularly encouraged to present their experiences with reverse debugging, memory debugging, non-interactive batch debugging, and long distance remote debugging.