IHPC's cluster will be used to provide more cost-effective computing solutions to its partners while SMA students will benefit from having direct access to
distributed computing solutions.
With a current staff strength of 170 research scientists and engineers, IHPC is well-placed to spearhead the use of advanced computational techniques and high
performance computing HPC resources across a broad range of disciplines, such as
manufacturing, electronics, chemical applications, precision engineering, data-mining, virtual reality, and advanced product design.
To provide such services, IHPC has invested and is investing millions of dollars
in SMP machines. Though well-built and stable, these shared memory systems are costly and are not easily scalable to meet the institute's growing needs. In
1998, the institute began experimenting with clustering PCs running on Linux. However, it will continue to use SMP systems due to its continued performance
and availability of commercial software.
"Our programmers were only experimenting with Linux for HPC then," recalled Dr.
Cheok Beng Teck, IHPC's director of plans and strategic development.
In November 2000, IHPC together with the Ministry of Manpower, secured a grant from the Enterprise Challenge Unit of the PS21 Office under the Prime Minister's
Office to develop the Collaborative One-stop Virtual Engineering System (COVES),
which would enable government agencies to tackle environmental, occupational health, fire safety, and emergency response planning issues during the design
stage of a factory. This would be achieved via the use of three-dimensional simulation of the dispersion of airborne contaminants, spills and ventilation.
"The biggest challenge for this project was to come up with a scalable computer architecture," said Dr. Cheok. "We would have no problems handling five plans
simultaneously, but should the number of plans increase ten-fold to 50 every week due to an investment building boom, our current systems would not be able
to process them fast enough. These are compute-intensive jobs."
The Institute purchased 30 IBM eServer xSeries x330 with Intel PIII 933MHz 1U nodes.
Running on Red Hat Linux, each node in the cluster has 1GB of memory and an 18GB
hard disk.
SMA, a major user of IHPC's SMP systems, also decided to purchase 70 nodes for training of students under the High Performance Computation for Engineered
Systems (HPCES) programme. "We are focused on training manpower for research and
education," said Prof Khoo Boo Cheong, Programme Chair, HPCES, SMA, which is a global partnership in graduate education between Massachusetts Institute of
Technology, the National University of Singapore, and Nanyang Technological University.
With the new Linux environment, IHPC is able to offer an additional more cost-effective computing platform. As a key component of its project charge with any
industry partner is based on cost of equipment, the overall rate to partners may
be lower due to the solution's lower price.
IBM had already announced that part of its Linux strategy would include investing more than US$200 million over four years in the Asia Pacific region
alone, to speed up the development of Linux-based products in Asia. In addition, investments in seven Linux development and competency centres in the
region and recruitment of some 300 consultants, researchers and product developers are also in the works.
Prof. Khoo said SMA's new installations would give students direct access to distributed computers for their masters and doctorate project work.
"Students will be attending courses that will help them use the cluster effectively in the shortest possible time" he added.
Collaborative relationship and efficient after-sales support Dr Cheok described the contract with IBM as a "collaborative
relationship." It includes a Memorandum of Understanding to promote Linux-based engineering
computations in the industry.