China to lead supercomputing sector
(China Business Weekly)
Updated: 2004-10-25 14:28

China will soon -- possibly next year -- be home to the world's fastest supercomputer, and the nation will hold all intellectual property rights (IPRs) to the bionic processor and its relevant applications.

That means China, previously a non-player in the advanced supercomputing that is the core technology to a country's information grid, soon will be among the world's top players in both supercomputing technology and applications, including the United States and Japan.

"We have already developed the world's fastest blade supercomputer," Steve Chen, founder and deputy chairman of Galactic Computing Shenzhen Co Ltd, said.

"What we will do is develop industrial applications based on the supercomputer, so that China will take the leadership in establishing key, strategic, nationwide information services grids." said Chen.

"We have contacted the government to apply for relevant projects in various industries. If things proceed well, we will launch our first product next year."

Galactic Computing, founded in 1999 with investment from Hong Kong-based Shell Electric Manufacturing Holdings Co Ltd, primarily develops blade supercomputers.

Chen, who has dedicated 30 years of his life to the supercomputing technology, is a member of the National Academy of Engineering of the United States and a member of the American Academy of Arts and Sciences.

"Our third-generation (blade supercomputer) is mature in technology and ready for deployment and development of industrial applications," Chen said.

The company's third-generation supercomputer, launched earlier this year, is reportedly the fastest computer of its kind -- real-time collaborative supercomputing blade system, in the world.

With an average calculation speed 1 teraflop (trillion floating-point operations per second), it can be scaled up to more than50 teraflops at peak speed.

In comparison, Chen noted, the blade supercomputer the United States is expected to launch, either late this year or early next year, will have an average speed of 41 teraflops. And Japan has announced that it will have a 50-teraflops blade supercomputer in 2005.

Chen last Saturday met senior Chinese Government officials from two dozen ministries and bureaus -- including education, commerce, public security, information industry, health, science and technology, and the National Development and Reform Commission.

Fastest computer in works

That marked the first time Chen met Chinese officials to seek government support for his blade supercomputing technology.

"We expect to receive favourable response from the government, universities and research institutions and companies ," Chen said.

He refused to say how much the projects would cost, but he did provide a reference point.

Chen said the US Government has invested US$90 million in the ongoing development of the 4-teraflop blade supercomputer, which was designed to assist with energy-related research.

"The cost will surely be somewhat lower in China," he said. "By far, we have invested US$20 million."

The US$20 million was funded by Galactic Computing's investor, Shell Electric.

Chen said his company will focus on 10 sectors when developing -- through co-operation with Chinese universities and/or research institutions -- industrial applications.

The 10 sectors will be healthcare, chip design foundry, education, entertainment, national security, logistics, new drug discovery and clinical trials, bioscience and exploration for natural resources.

"We have selected eight top colleges or research institutions as development partners, each for a specific industrial application," he said.

For example, Tsinghua University was selected for bioscience; China University of Geosciences, exploration of natural resources; Beijing Jiaotong University, transportation logistics; and Nankai University, port logistics, Chen explained.

Galactic Computing, with its partners, will invite the leading Chinese company in each sector to adopt, when ready, and promote the relevant application, Chen said.

"With our technology transfer, I hope we can help China in making a leapfrog in blade supercomputing technology and applications, and joining the global leaders in information services economics,," Chen said.

China lags behind other nations, including the United States and Japan, in supercomputing technology and applications, he added.

Dawning Information Industry Co Ltd, a leading Chinese supercomputer vendor, launched, earlier this year, a 1-teraflop supercomputer. To date, that is the fastest supercomputer in China.

Yet despite the notable progress, China does not have much IPRs to the computer's core system, which is based on imported technology, Chen said.

Also, the design of Dawning's product was not based on the "blade," which has become an accepted concept in supercomputer design, he noted.

Some major computer vendors in the United States, aware of the importance of blade supercomputing system, are starting research in the field.

Yet, "they are still in the early stages, and are focusing on the hardware. We started six years ago. Now, we are prepared to develop applications," Chen said.

"I decided to move our work from the United States to China, as I believed the products and the relevant applications would be better promoted here," he said.

The United States, despite being a leader in super computer technology, was not overly interested in adopting the blade design when Chen invented the concept.

That was because the US Government had spent so much on-speed-oriented supercomputers, and that heavy investment made it hard for US officials to shift their focus to a new technology, Chen said.

Despite the fact many US companies -- including IT (information technology) giants such as IBM and Dell -- immediately recognized the design as the technology trend of supercomputers, "they would not promote the technology as we do, as it is against their box sales business models," Chen added.

Users of blade supercomputers need not discard old computers to buy new ones for greater performance or new applications, Chen said.

Instead, they can simply re-use the old blades and add new "blades" to the existing equipment. That, Chen added, creates savings.

"The technology is not welcomed by computer vendors as it does not require frequent purchases of entire systems," Chen said. "We are not simply a vendor."

In comparison, China, the latecomer, has few obstacles in adopting the new technology.

"Companies or institutions do not need to buy such supercomputersand use it expensively after the government pays the bill prior to establishing Galactic Computing, Chen worked in the United States since early 1980s.

He created Supercomputer Systems Inc in 1987, with financial aid from IBM, after he left his post as a vice-president of renowned US supercomputer company Cray Inc.

"It was difficult to import supercomputers into China in 1980s. Now the component technologies for commercial use, becoming more mature, are available for us to build high-performance supercomputers for China market ," he said.

"But our ultimate goal is not only China. China, as a leader in the near future, should export its blade supercomputing technology worldwide," Chen said.

The idea of blade supercomputing, first raised by Chen in 1998, is widely considered to be a revolutionary design concept. It has significantly enhanced the sustainability of computers' processing capabilities.

Users can easily increase the calculation speed of a blade supercomputer by adding "blades," or sets of CPUs (central processing units).

The different computing sections of a blade supercomputer can achieve real-time, collaborative computing in a single system, Chen explained. While in traditionally-designed supercomputers, the processing capability is fixed and each computing section is programmed for a certain function, working independently.

According to Chen's scheme, China needs 50-100 units of huge blade supercomputers, to work as a systematic public information service platform covering diversified sectors.

The existing platforms, most of which were set up one by one, scattered and isolated from each other, will be replaced.

"It is similar to the electric power utility industry, when a few large nuclear power plants are set up to replace numerous small power generators... it is much more effective and efficient," Chen said.

Close