In this video, Robin Humble, NCINF, presents Lustre as a root file system. Recorded at the LUG’2011 conference in Orlando on April 13, 2011. Slides from the conference will be posted soon at the conference site.
Video: Lustre as a Root File System / Speeding up Metadata
Link and run for 08/05/2010
Japanese Supercomputer Ranked 1st In The Little Green500 List
A Japanese university and research institute has announced that their supercomputer system was ranked first in a ranking of supercomputer’s performance per unit power consumption. The supercomputer system, Grape-DR, was ranked first in the Little Green500 list, getting a 5% higher score than IBM Corp’s system that is located in Germany, and ranked second in the list. In order to achieve this, they combined 64 pairs of Intel’s Core i7-920 processors and a board mounted with four of the Grape-DR accelerator chips.
Mathematics + Supercomputers = Big Bang Explained
Scientists now have a better chance of finding answers to that mystery because of the massive computational power of supercomputers – today’s fastest, most powerful computers, says Daniel R. Reynolds, assistant professor of mathematics in Dedman College.
CHPC Awards First GPGPU Cluster Contract to Orange Business Services
Orange Business Services has been awarded the supply and implementation order for the first GPGPU (general-purpose computation on graphics processing units) high performance computing cluster at the Centre for High Performance Computing (CHPC) in South Africa.
BioTeam Tweaks Amazon’s HPC Cloud
The BioTeam has devised a method for expanding the boot volume of Amazon’s HPC-oriented Cluster Compute instances. The guys at BioTeam wanted to run bonnie++ (a benchmark suite for disks and filesystems) on an Amazon boot volume but Bonnie++ ideally needs to work with a system that is roughly twice the size of the currently available 23GB of physical RAM on the Cluster Compute instances, which also only has 20GB of disk space. However after 20 or so minutes of tinkering (by really technically endowed people), they claim that they were able to expand the system disk to 80GB. The pleasant surprise here according to BioTeam is that all of this was possible without having to debug boot failures or grapple with ugly kernel panics or any of the other issues that one runs into when partitioning low level disk systems.
Linux supercomputer, worth £2m, sought by University of Warwick
The University of Warwick is tendering for a new Linux-based High Performance Computing facility for its research Centre for Scientific Computing (CSC). A “significant” share of the new facility will be used for research in the field of magnetohydrodynamics (MHD – the study of the dynamics of electrically conducting fluids such as plasma and metal liquids), to support the computational requirements of the UK MHD research community. The facility will also be used to support research from other disciplines at the university.
Budding Engineers Build a Supercomputer
From 6th to 9th July, twenty 16 and 17 year-old students descended on the University of Southampton to take part in a Supercomputing course put together by The Smallpeice Trust and delivered in partnership with the University of Southampton’s School of Engineering Sciences and Microsoft.
Link and Run for 07/09/2010
EMC to Acquire Greenplum
EMC Corporation, the world’s leading provider of information infrastructure solutions, today announced it has signed a definitive agreement to acquire California-based Greenplum, Inc. Greenplum is a privately-held, fast-growing provider of disruptive data warehousing technology, a key enabler of “big data” clouds and self-service analytics. Upon completion of the acquisition, Greenplum will form the foundation of a new data computing product division within EMC’s Information Infrastructure business.
Heterogeneous Multi-Core Chips Might Be Needed for Exascale Computers
The next big thing for supercomputers are projected to be exascale machines. The leading chip designers are working on technologies that will enable the next leap in the high-performance computing space. According to an HPC expert from the University of Tennessee, in exascale systems will require new central processors, graphics processors or hybrids that combine both onto the same piece of silicon. But, Cell chips, which are heterogeneous multi-core processors, are dead end.
Introducing OpenCL
Over the past decade, graphics cards have gone from being simple accelerators to being fast general-purpose computing engines. David Chisnall looks at OpenCL, a new API for running non-graphics applications on modern GPUs.
Australians launch new Compute Cloud
With the launch of their new Compute Cloud last month, the Australian Research Collaboration Service aims to bring user friendly grid and cloud computing to a whole new level.
Supercomputer centre to boost telescope bid
The Parliamentary Standing Committee on Public Works yesterday recommended the Federal Government spend $66 million to build the Pawsey High Performance Computing Centre in Perth.
Link and Run for 06/25/2010
Do Processors Really Matter Anymore?
But now that processors have gotten so powerful and the ability to pool them has been simplified, is it time to start wondering whether the old ways of placing value on new equipment are no longer valid?
Science at Scale: SciDAC Astrophysics Code Scales to Over 200K Processors
MAESTRO, a low Mach number code for studying the pre-ignition phase of Type Ia supernovae, as well as other stellar convective phenomena, has just been demonstrated to scale to almost 100,000 processors on the Cray XT5 supercomputer “Jaguar” at the Oak Ridge Leadership Computing Facility. And CASTRO, a general compressible astrophysics radiation/ hydrodynamics code which handles the explosion itself, now scales to over 200,000 processors on Jaguar—almost the entire machine. Both scaling studies simulated a pre-explosion white dwarf with a realistic stellar equation of state and self-gravity.
Parallel Studio Meets Visual Studio 2010
Intel’s James Reinders talks with Steve Teixeira, Product Unit Manager for Parallel Developer Tools at Microsoft, about the new Visual Studio’s support for parallel programming.
Supercomputer helps telescope see echos from the big bang
An international collaboration of scientists has announced the first results of the ACT project, probing the early years of the Universe, at Canada’s largest supercomputing conference in Toronto today. The presentation was made by Jonathan Sievers, of the Canadian Institute for Theoretical Astrophysics.
Link and Run for 06/22/10
PBS Works 10.4 Increases Accuracy and Predictability for HPC Capacity Planning and Forecasting
The new version of PBS Professional®, the foundation for PBS Works, offers customers a next-generation backfill scheduling system with the ability to accurately predict job start times. PBS Analytics™ 10.4 introduces a new standalone resource utilization data analysis and visualization module, License Analytics. License Analytics can be deployed within any IT environment and is an easy-to-use solution that supports data-driven planning and decision-making. Together, the enhancements allow customers to maximize information technology (IT) investments by enabling more accurate forecasting and capacity planning decisions.
Chelsio Announces Unified Manager 1.1
Unified Manager centrally manages all Chelsio network adapter cards on the network across Windows and Linux environments. It includes an intuitive GUI and command line interface, providing a simple to use, fast tool for IT managers to access every iSCSI initiator or target, iWARP peer, TCP/IP Offload (TOE) enabled and stateless NICs on a network. Unified Manager now works on a wide range of operating systems: Windows 2003 and 2008, RHEL 4, 5, SLES 10, 11, Solaris 10, Open Solaris, FreeBSD and Mac OS.
Voltaire Announces New Low-Latency Layer 2/3 Switch
The combination of the Voltaire Vantage 8500 Layer 2 core switches and new Vantage 6024 switches enables customers to build flat datacenter fabrics of more than 3,400 10 Gigabit Ethernet ports with non-blocking, lossless switch capacity of 69.12 Terabits per second. Voltaire Unified Fabric Manager (UFM) software orchestrates the fabric as a single logical entity, enforcing fabric-wide service policies, providing real-time fabric and application level monitoring, and simplifying fabric administration across many physical and virtual switching elements.
Link and Run for 06/18/2010
Microsoft begins expansion of its Data Center in Quincy, Washington
Earlier this month, Microsoft began a major expansion of its Columbia Data Center in Quincy, Washington and has started to excavate ground for the ‘Phase 2’ data center building (see attached photos of the excavation at the Quincy site).
This Week’s Multicore Reading List
A list of book releases compiled by Dr. Dobb’s to keep you up-to-date on parallel programming and multicore technology.
SGI Releases InfiniteStorage 5000 SAS External Storage System
New Storage Product Delivers 6 Gb/s SAS Technology Offering Mid-Range Performance, Scalability and Features at Entry-Level Price Points
Appro Deploys a World Class Linux Cluster Testbed Solution to LLNL in Support of the Hyperion Project
Appro, a leading provider of supercomputing solutions, today announces the deployment of a Data Intensive Testbed Cluster solution based on Appro servers configured with ioMemory technology from Fusion-io to extend the existing Lawrence Livermore National Laboratory (LLNL) Hyperion system. Hyperion is the world’s largest Linux cluster dedicated test bed for development and testing of critical Linux cluster technologies.
SGI Helps Queensland Government Accelerate Climate Science In Australia
SGI, a global leader in HPC and data center solutions, today announced that the Queensland Department of Environment and Resource Management (DERM) has installed a high performance computing (HPC) and large-scale storage solution based on SGI Altix ICE 8200 and InfiniteStorage in its Brisbane facility for the processing of computationally intensive environmental modeling applications.
LSU Physics Professor Jorge Pullin Chairs Selection Committee for Einstein Prize
Jorge Pullin, a professor in the LSU Department of Physics & Astronomy and Interim co-Director of the LSU Center for Computation & Technology, or CCT, is chairing the American Physical Society’s selection committee for the prestigious Einstein Prize.
Link and Run for 05/29/2010
OpenCL Optimization Case Study: Diagonal Sparse Matrix Vector Multiplication
This article discusses performance optimizations for AMD GPUs and CPUs using as a case study a simple, yet widely used computationally intensive kernel: Diagonal Sparse Matrix Vector Multiplication. We look at several topics which come up during OpenCL performance optimization and apply them to our case study.
3-D model of blood flow by supercomputer predicts heart attacks
EPFL Laboratory of Multiscale Modeling of Materials, in Switzerland, has developed a flowing 3D model of the cardiovascular system that should allow for predictions of certain heart diseases before they become dangerous.
Processor Whispers — About Partnerships and Partner Chips
The IT economy is doing better and better, despite the renewed spitting – and now even cloudy twittering – of the Icelandic volcano Eyjafjallajökull. This positive development could very well lead to new partnerships and almost-forsaken projects like Intel’s Larrabee graphics chip might make a comeback.
CSC makes $317M play for NOAA supercomputer
Now that hurricane season 2010 is open, there will be plenty of talk of severe weather in the coming months. And whether it is tornadoes, heat waves or floods there is little that can be done about it. But improved forecasting would go a long way toward ameliorating some of the more devastating effects. That’s a big reason why the National Oceanic and Atmospheric Administration’s mission, signed Computer Sciences Corp. to a nine-year $317 million to build a supercomputer for modeling weather patterns.
SGI advances Linux on the HPC front
Perhaps it is this acceptance of open source that led SGI’s CTO, Dr Eng Lim Goh to ditch a fancy conference setting and instead talk with a bunch of technologists at a Greater London Linux User Group (GLLUG) meeting held at University College London (UCL). Goh’s belief in his firm’s technology and that of the open source community was given away by the title of his talk, “Linux is supercomputing”. Underlying that claim was the fact that SGI has managed to get the standard Linux kernel, available from the kernel.org repository, to work on systems having 4,096 cores.
Researchers Aim to Achieve Cleaner Coal Through Computation
It’s not possible to see inside the flue where the gases are interacting, so Wilcox simulates the interactions of these particles using the Ranger supercomputer at the Texas Advanced Computing Center (TACC). Her studies of the dynamics of trace metals inside the flue of a power plant are helping her design and improve the technologies capable of removing heavy metals from the combustion process.
Hutchinson Center Receives $10.1M for HPC Cluster and Datacenter
Fred Hutchinson Cancer Research Center has received two grants totaling $10.1 million from the National Institutes of Health to fund a new high-performance computing cluster and the creation of a campus-based facility to consolidate and safeguard research data.
Fujitsu Supercomputer Achieves World Record in Computational Quantum Chemistry
Fujitsu Limited and Chuo University of Japan today announced that a team of researchers from Chuo University, Kyoto University, Tokyo Institute of Technology and Japan’s Institute of Physical and Chemical Research (known as Riken) employed the T2K Open Supercomputer — which was delivered by Fujitsu to Kyoto University’s Academic Center for Computing and Media Studies — to successfully compute with high precision, as a world first, an optimization problem to reveal the molecular behavior of ethane (CH3 only), ammonia (NH3) and oxygen (O2).
Research Grant Link and Run
A quick series of links from a few institutions garnering funding for high performance computing activities.
The Fred Hutchinson Cancer Research Center Receives $10 Million Grant
The Fred Hutchinson Cancer Research Center said today it has received two grants worth a total of $10.1 million to set up a high-performance computing cluster and center to safeguard research data. The money comes from the American Recovery and Reinvestment Act, aka “the stimulus,” via the National Institutes of Health. The plan is to create an energy efficient facility that will safely store “irreplaceable research data,” and to buy high-powered instruments, the center said in an e-mailed statement.
Lehigh Technologies, an innovative manufacturer of high-performance, engineered rubber powder, announced it has been awarded a National Science Foundation (NSF) grant to support research into the development of commercially viable composites of plastics and micron scale engineered rubber powders manufactured from end of life tire materials. The research grant, under NSF’s Small Business Innovation Research (SBIR) program, will fund a collaborative program involving Lehigh Technologies scientists and a team from the Georgia Institute of Technology.
Link and Run for 05/21/2010
Each week I read (well, skim) a bunch of posts. Many of them are clearly uninteresting to you, my discriminating reader (although the firings at the Health Professions Council are sometimes quite interesting). Some of what I read is both interesting and relevant, and those get read in detail and posted, sometimes with quite a bit of analysis from me.
Then there is the stuff in the middle: might be interesting to some of you, but either there isn’t much more to say than what’s in the post I’ve found or I don’t have time to write it up. Until today I’ve just thrown those links away.
Today that carnage of content ends! Starting today I’ll be posting at least some of the links that I think are interesting but for whatever reason don’t be written up in these “Link and Run” posts. Look for them about once a week.
Infusion adds HPC competency
“Infusion’s deep software development and training expertise will help customers in financial services and government take full advantage of Windows HPC Server 2008, including new capabilities that bring supercomputing power to Excel 2010,” said Vince Mendillo, senior director, Microsoft High Performance Computing Group.
As part of announcing this relationship, Infusion will begin conducting Discovery Workshops in North America, Canada and the UK. These Discovery Workshops are centred on HPC services for Excel 2010 and targeted towards financial services customers.
Dell taps ARM processors for low-power servers
Dell Inc. is set to test multicore ARM processors from Marvell Technology Group for possible use in low-power servers for large data centers. The company has already shipped a few thousand low-power servers based on x86 processors from Taiwan’s Via Technologies Inc.
SGI to distribute LSI enterprise storage products
SGI, a provider of HPC and data centre solutions, has agreed a global initiative with LSI Corporation to expand its SGI InfiniteStorage products with new configurations powered by LSI. This expanded relationship will make available a wider breadth of LSI’s storage technology through direct and indirect sales force.
Vector Fabrics cloud tools to automate sequential C code analysis for multi-threading
Vector Fabrics announces vfAnalyst, a cloud-based tool for parallelizing sequential C code. The first in a planned family of cloud computing tools, vfAnalyst enables software engineers to easily identify the most promising parallelization opportunities so that they can create an effective multicore implementation much more quickly than is possible today.
Structured Parallel Programming with Deterministic Patterns
This document discusses and advocates a structured approach to parallel programming. This approach is based on a core set of common and composable patterns of parallel computation and data management with an emphasis on determinism and scalability. By using these patterns and also paying attention to a small number of factors in algorithm design (such as data locality), programs built using this approach have the potential to perform and scale well on a variety of different parallel computer architectures.
