| Magazine Issue 27 June 2003 |
|
Advertisement
|
 |
| |
| Start |
|
|
| A new design for supercomputers? |
|
Clusters built of commodity processors are invading the TOP500. In the current list, they reach as high as postition 3. Hans Meuer, one of the list compilers, expects that they could even dominate the list in 2 years time. On the other hand, Horst Simon presented evidence that the current off-the-shelf approach has reached its limits. In a presentation the German company Acconovis showed considerable speed-up using special processors. Also some of the other exhibitors showed this approach. And, of course, Cray and NEC still produce proprietary chips.
For the other news form the major HPC event in Europe, please read the two PrimeurLive! issues for more information:
PrimeurLive! Issue June 25 2003
PrimeurLive! Issue June 27 2003
Please do not forget to check out our regular Primeur Magazine issues that keep you up to date with all HPC and Grid activities.
|
|
|
|
Advertisement
|
|
| |
| Focus |
|
|
| GRIA takes Grid computing into the real world |
|
In the Dolphin booth at the ISC 2003 Conference, Atle Vesterkjaer presented a demo to PrimeurLive! showing how Grid technology can be applied in commercial settings. Mr. Vesterkjaer demonstrated a first version of the application developed within the GRIA project. GRIA stands for GRID Resources for Industrial Applications. The aim of GRIA is to enable industrial users to trade computational resources on a commercial basis to meet their needs for more computer power in a cost-effectively way.
Read further... |
|
|
|
| It is hard work to keep up with people expecting us to follow Moore's law |
|
Hardware designers are fulfilling the expectations of everyone to follow Moore's law for many years. They do not realise that in fact it takes hard work from the chip designers to keep up with the expectations. In Heidelberg we had a short interview with Jerry Huck, one of the main Itanium chip designers at HP.
Read further... |
|
|
|
|
|
| |
| TOP500 supercomputing |
|
|
| Off-the-shelf supercomputing is a dead end |
|
Currently most US supercomputers are built from off-the-shelf components. Evidence in 2002 piled up that this is a dead end, explained Horst Simon, director of NERSC at ISC2003 in Heidelberg. First there was the Earth Simulator, that acted like a Compunetik to the US supercomputing community, then there was a workshop on Petarange supercomputing, that showed no progress was made in supercomputing architecture during the past five years, and then there were the poor benchmark results on the newly available supercomputers. Supercomputers need to be designed with scientific applications in mind. First examples of co-operations towards this approach are "Red Storm" and "Blue Planet".
Read further... |
|
|
|
| Interdependence of architecture and software for effective terascale computing |
|
Dr. Thomas Sterling, California Institute of Technology, highlighted the important relationship between computer architecture and the software to achieve an effective computing in the terascale range. The performance, performance to cost, efficiency, and programmability as well as the opportunities afforded by technology advances are demanding simultaneous and strongly interdependent innovation in system architecture and software. He discussed the potential and nature of these interdependencies and the innovation they will create.
Read further... |
|
|
|
| Building a PetaFlops class machine for large scale system design experience and biomolecular simulation |
|
Mr. Jamshed Mirza from the Deep Computing Department at IBM addressed the painstaking issue of the laborious emergence of Pflop/s (Petaflops) systems in relationship to the BlueGene initiative at IBM. The road to achieve Petaflop computing power has shown itself to be a tentalising one: the goal always seemed within sight but always stayed just out of reach up till now because of the multiple challenges and potholes on the way, as Mr. Mirza expressed it.
Read further... |
|
|
|
| Exploring the benefits of FPGA-processor technology for genome analysis at Acconovis |
|
At the ISC 2003 Conference, Holger Singpiel from Acconovis in Mannheim and Carsten Jacobi from BASF in Ludwigshafen presented the FPGA technology and its use in high throughput genome analysis. FPGA processor-based genome research offers promising results with regard to speed and accuracy in the analysis and modelling of complex energy models.
Read further... |
|
|
|
| Twenty years experience at NAL with software for HPC in aerospace science and engineering |
|
Yuichi Matsuo, National Aerospace Laboratory (NAL) of Japan, gave a brief overview of the computing history like the Numerical Wind Tunnel and presented requirements for the new Numerical Simulator (NS). As can be seen from the Top500 list NAL chose a cluster of Fujitsu Primepower systems. Then he discussed first experiences and expectations with regard to the new machine.
Read further... |
|
|
|
| Software for large-scale computing: it is scalability that matters! |
|
Lutz Gross, Australian Computational Earth System Simulator, University of Queensland Brisbane, considers scalability as the most important factor. He presented some examples concerning algorithms, problem size and possible solutions.
Read further... |
|
|
|
| Can SuperData Centres be secured? |
|
The main topics of this question were discussed by Brian Koch, RxHUB, Saint Paul, USA. He presented the problems of openness of the huge academic centres compared with governmental and industrial secured computer centres. He looked at those centres through a security lens and saw it as an acceptable or tolerable level of risk.
Read further... |
|
|
|
| Complexity of data in the passenger services systems of the DB AG |
|
Lutz Philipp, DB Systems, explained different processes of the Passenger Services of the Deutsche Bahn AG in theory and practice. Presenting on behalf of Karl-Heinz Holzwarth, who could not make it to the conference, he emphasised the development of the data processing systems of the DB AG from a historical point of view, the increasing complexity of data, characterised by the data complexity with the DB AG, the problems in relation to data complexity, presented practical examples and looked towards future developments.
Read further... |
|
|
|
| Billing of million customers at German Telekom |
|
Christoph Böhm, Telekom AG, Bonn, presented evolutionary strategies for profitable high volume customer data handling in the view of German Telekom. He described the most important issues of billing, the billing process and methods and the challenges and issues.
Read further... |
|
|
|
Advertisement
|
|
| |
| The Grid |
|
|
| Taming huge data volumes |
|
Sverre Jarp, Cern, discussed the topic "Storing and Processing of Huge Experimental Data at CERN". There he described the requirements of High Energy Physics, the computing characteristics and the planning of storing, distibuting and processing of the huge data volumes, the Large Hadron Collider will deliver when he will be active.
Read further... |
|
|
|
Advertisement
|
|
| |
| Company news |
|
|
| Rapidly evolving microprocessor technology turns throughput computing into alternative for HPC |
|
Marc Tremblay, Vice-President and Fellow at Sun Microsystems, addressed the relationship between throughput computing and HPC at the annual ISC Conference in Heidelberg. Throughput computing will definitely change the cost of computing and the emerging demand for high bandwidth will approach throughput computing-based commercial systems to the ideal of smoothly functioning HPC systems. Sun is able to deliver throughput computing via CMT/Sparc, 64-bit, Solaris and software stack in both commercial and HPC markets. These were the statements Mr. Tremblay made in front of the ISC 2003 audience.
Read further... |
|
|
|
| Dell introduces 64-Bit server for high-performance computing market |
|
Dell has introduced the PowerEdge 3250 server, a high-performance 64-bit system designed to address sophisticated computing requirements such as advanced graphical rendering in special effects, complex financial modelling and genomic research.
Read further... |
|
|
|
| Efficient network-storage, TCP processing and processor development under the loop at Intel |
|
Justin Rattner, Intel Senior Fellow and Director of Microprocessor Research at Intel Corporation, gave a key note lecture at ISC 2003 on how to build efficient HPC systems from catalogue components. Mr. Rattner believes the economics of high performance computing underwent a major change. HPC solutions need to track Moore's Law in order to remain viable. Mr. Rattner is also convinced that Intel is playing a key role in accelerating HPC tools for science, engineering and business with open commercial off-the-shelf technology expertise.
Read further... |
|
|
|
| AMD Opteron processor answer to tough challenges in high performance computing |
|
Michael Goddard from AMD introduced the new AMD Opteron processor to the attendees of the ISC 2003 Conference. As detailed as he was on the AMD Opteron processor features, as little he would expand on performance and software infrastructure. Let us have a look though at what Mr. Goddard had to say about the six challenges that advanced multiprocessor systems have to deal with.
Read further... |
|
|
