News

The Business Times
By TANG WENG FAI

SPURRED by interest in Linux by enterprises around the region, mainboard-listed Singapore Computer Systems (SCS) has spun off an independent unit to exploit opportunities in supporting the free operating system.

SCS set up Scalable Systems last November to provide state-of-the-art high performance cluster computing (HPCC) solutions to the growing grid and cluster computing markets worldwide. These solutions use exclusively Linux to build low-cost computing clusters made up of many computers working together to perform as a supercomputer.

Major shareholders include Singapore Technologies Incubators@Work Pte Ltd, SCS Ltd and two leading figures in the local HPCC community who are also staff founder members.

'We're seeing real interest and money, though small (now), in the region for enterprise scale support for Linux. This isn't a fad,' Laurence Liew, chief technical officer for Scalable System told BizIT.

That Linux is increasingly flexing its muscles in the enterprise world is evident from the latest Top 500 list of supercomputers released late last November by the University of Mannheim and the University of Tennessee. A decade ago, the list was dominated by traditional mainframes that ran dedicated operating systems. Today, about half of the systems on this list are running clusters built from commodity hardware and most of these are using the Linux operating system.

'You can now start building a respectable supercomputer on budgets less than US$500,000, which was unimaginable 10 years ago,' said Mr Liew.

In Singapore, systems that SCS/Scalable Systems have been responsible for building include a 128-node Intel-based Linux cluster for the National University of Singapore's (NUS) School of Computing.

Another two clusters at NUS, 60-node and 32-node, for the NUS-MIT Alliance and the Supercomputing and Visualisation Unit respectively, were also implemented by SCS.

BizIT understands that another such cluster at the Nanyang Technological University has also been implemented by the SCS/Scalable Systems team.

The team is working with the US-based San Diego Supercomputer Center (SDSC) and the University of California, Berkeley, in the development and support of a 'Rocks Cluster' tool kit. The software allows many nodes of a large Linux cluster to be efficiently linked together to perform a given task.

The latest version of the tool kit was released on Dec 23 last year, and now supports three common microprocessor families; the Intel IA-32, the Intel Itanium and the AMD Opteron.

An enhanced version is to be commercially released by Scalable in early 2004.

The regional push is given added credibility by separate MOUs signed with Dell Computer Asia and Intel Technology Asia. Under the terms of the respective MOUs, Scalable Systems will offer the Scalable Rocks Cluster Toolkit on Intel Xeon and Itanium processor-based Dell systems in the Asia-Pacific region.

Initial markets targeted are Singapore, Malaysia, Thailand, Australia and New Zealand as these are the neighbouring markets identified as having significant cluster computing market potential and not requiring language localisation. Of these, Australia has been identified as currently having the most HPCC market potential.

'The North Asia markets (Japan, China, Taiwan and Korea) are also under consideration, but language support and security issues will need to be worked out first,' says Mr Liew. Japan is seen as having the greatest HPCC market potential in the Asia-Pacific region.