logo
EnterTheGrid - Primeur Live!

EnterTheGrid - Primeur is the premier Grid and Supercomputing information source in the world. With Primeur Live! it brings you Live reports from Europe's main Supercomputing and Grid events

>Primeur Magazine
>PrimeurLive!
>EnterTheGrid
>Analysis
>Backissues
>Calendar
>Subscribe
>Advertise
>Contact
Issue 25 June 2003
>Start
>More participants at ISC2003 supercomputing conference in Heidelberg
>Focus
>ALiCE Cluster in operation for 3 years
>TOP500 supercomputing
>Clusters march into TOP3 of the TOP500 supercomputer list
>Europe losing ground in the TOP500
>A 100 Tflop/s supercomputer for the UK in 2005
>The Earth Simulator evaluated after its first year in business
>Selecting the best suitable HPC architecture for Earth system modelling at ECMWF
>Community Climate System Model to simulate ocean, land and atmospheric models
>The future of anthropogenetic, historical and evolutive climate change research in Europe
>The Grid
>On-line Science the worldwide telescope as a prototype for the new computational science
>HPC and Grids in Asia
>Company news
> SuSE Linux enterprise server selected by Cray for Department of Energy's new "Red Storm" supercomputer
>First 15 nodes of the NEC SX-6 supercomputer installed successfully at UK Met Office in Exeter
>ClusterVision to install first supercomputer cluster in Europe based on Infiniband technology
>Intel, Swiss Institute of Bioinformatics and HP open Life Sciences Center in the Swiss Biotech Valley
HPC and Grids in Asia
Heidelberg 25 June 2003 At the Heidelberg Supercomputer Conference, Kahaner, from ATIP in Japan, gave a "virtual presentation" to the audience on HPC in Asia, answering questions over a link to New York. In Asia, in the high-end HPC, the Earth Simulator is the major achievement, that creates new science. But the USA breadth and depth in hardware, software applications, and system integration, will continue to provide leadership. In the "real world" in Asia, below the level of the Earth Simulator, there is significant creation of capacity, software and system integration developing locally in most countries.
Advertisement
Dolphin's SCI interconnect features the lowest latency and wire speed
Advertisement
Visit our sponsors

In Japan, hardware developments are mostly financed indirectly through government procurements for large supercomputers in universities and research institutes. The development of new large systems is hostorically linked to major procurements at tha National Aerospace Laboratory (NAL) and the Japan Atomic Energy Research Institute (JAERI). However, according to Kahaner, the future funding through government procurement is increasingly uncertain due to university reforms and tighter control of the governemn R&D budget. The focus is less and less on processor development. In academic and research, the focus today is on Grids an bio-tech applications.

The Earth Simulator seems to be a unique system. Kahaner does not see momentum in Japan to create a follow-on. New science on the Earth Simulator that is already coming out, includes advance in electronics. Strange as it may seem, the buzz on the Earth Simulator is mostly from outside Japan. Inside people are interested, but the Japanese emphasis is more on applications, not on hardware.

In the USA, however, the Earth Simulator has renewed interest in HPC, sparking off several programmes. All in all the system has a positive impact on the HPC community worldwide.

Several special purpose machines are under development in Japan, including the 2 Pflop/s (Peta-flop/s) Protein Explorer at the Genomics Sciences Center and the embedded eHPC special purpose LSI to be used for physics, chemistry and nanotechnology research.

Japan has also serious plans for large clusters, including the AIST SuperCluster in Tsukuba, that is targetd at 10 29 Tflop/s performance. At CNRC already a 1064 cluster is installed.

Notwithstanding energetic discussions about their efficiency, Kahaner sees grids everywhere in Asia. Most countries have their national Grid projects. In Japan, there are for instance the Tokyo Institute of Technology Grid, the Osaka Life Sciences Grid and the Riken GSC Life Sciences Grid.

In 2003 projects were started for the development of Business Grid, designed to enable Japanes vendors to compete wiht IBM Grid developments and NaReGi, the National Grid Research Initative. And there is the Grid Technology support centre (AIST) at Tsukuba.

In Asia outside Japan, one is looking to the USA for inspiration, systems and software. There is not any fundamental system development, but, Kahaner said, there are many examples of COTS-systems. In China, there is for instance, Legend, and in India, the CDAC PARAM system. HPC is driven by Bioscience needs in most Asian countries.

In China, US vendors still dominate the market, but there are three Chinese companies that are very active: Dawning, Legend and LangChao. The Acadmy of Mathematics and Systems Science in Beijing, for instance, has a 2 Tflop/s legend system that consists of 512 Xeon processors with Myrinet interconnect.

China has a national Grid Project that runs from 2002 to 2005. The Chinese Academy of Sciences operates an e-Science Grid and there is also an Education Grid.

The China National Grid consists of some 10 nodes with an aggregrate performance of 6 - 10 Tflop/s, mainly Dawning Grid enabled clusters, but also including Sun, SGI and Sunwei systems.

According to Kahaner, China believes that HPC and Grid technology will reach mass adoption stage in less than 15 years. Two of the largest markets will be China and India. Grid offers opportunities for innovation.

In India, there were several parallel systems developed in the 1990s, but most have gone away, Kahaner said. The only one left is the Center for Development of Advanced Computing (CDAC). But this is currently doing more in developing system software and applications for commercial systems. The PARAM 1000 for instance, is a cluster of Sun Ultra e450 workstations with Fast Ethernet, Myrinet and the internal develped PARAMNet.

THE PARAM Padma has compute nodes based on Power4 RISC processors, in an SMP configuration. Nodes are connected by PARAMNet-II. The Padma runs C-DAC's flexile and scalable HPC software environment. It uses Fibrechannel to connect Raid storage servers, SAN, and parallel file servers. It also includes hierarchical storage management technology.

India also has collaborations with Russia in the Russian Indian Center for Advanced Computing Research (RICCR), a collaboration between C-DAC and ICAD in Moscow, including applications in Aerohydrodynamics, Atmosphere and Ocean research, Medicine and other application areas.

Advertisement
Advertisement
Visit our sponsors
Ad Emmen

EnterTheGrid - Primeur

James Stewartstraat 248

1325 JN Almere

The Netherlands

http://EnterTheGrid.com

mailto:primeur@hoise.com

© EnterTheGrid - Primeur Live!