Need for a new grid

It's a chicken and egg dilemma for the proponents of grid computing in Singapore.

By Alfred Siew, March 13, 2002

Grid computing, the latest computing buzzword in town, is turning out to be a complex challenge for its proponents here.

The technology involves linking clusters of powerful computers in remote locations so researchers can tap on their resources when-ever they need to perform massive computational tasks.

Advertisement

"Imagine plugging your notebook into a port like you would into a network and getting the power of a cluster of computers," said Associate Professor Teo Yong Meng of the National University of Singapore's (NUS) School of Computing.

However, building an extensive grid, like the Biomedical Grid, may require the upgrading of existing network infrastructure.

Grid computing will enable specialist jobs such as bio-medical research become faster and more efficeint.

This is because the computer clusters at various research institutes have to be connected with high-speed links.

Less than 10 years after the high-speed Singapore ONE network was set up, there is now talk among experts that a faster national network is required to support the Biomedical Grid.

Singapore ONE maxes out at 655 Mbps (megabits per second) now, while SingTel and StarHub can potentially offer up to 2.5 Gbps (gigabits per second) with their network equipment.

But a network of 10 to 40 Gbps - like ones being built in the US and Canada - are what researchers desire.

"If it's too slow, it doesn't make sense to send a job to a grid (network). You might as well do it locally," said Associate Professor Lee Bu Sung of Nanyang Technological University's School of Computer Engineering Building a fast network - possibly one involving optical switches - comes with cost, naturally.

"It's a bit early for a 10- GB metropolitan network here - there's no commercial case now," said Mr Mock Pak Lum, 1-Net Singapore's chief executive officer. "It may be easier if we can get the Government to fund the infrastructure, just like the MRT network."

Another concern among experts is interoperability across computing platforms. Despite the recent buzz, Assoc Prof Lee said that Singapore is moving at a conservative pace.

What is needed is improved management software to broker jobs and bill those who want to use the resources, he said.

One answer is the ALiCE (Adaptive and scalable Internet-based Computing Environment) project led by Assoc Prof Teo.

The case for a national grid is the insatiable desire for data churning, especially with the Government's plan to build a biomedical hub here.

Now it's a matter of how much resources will be pumped in. More money and researchers can expect faster machines and networks on the Biomedical Grid, said those involved in the project.

Said Assoc Prof Teo: "Do you build the highway first, or do you wait for the traffic? I think there'll be no traffic without the highway."

Crunching data

Grid computing involves pooling computing resources so they can be tapped by anyone who needs to crunch data, be it for simulation, biomedical research or image rendering.

It builds on the premise of idle computers which do little outside office hours. For example, a bank here may send a complex calculation job to clusters of computers in the US, where the offices are closed at night.

This way, the bank does not have to build and maintain an expensive computer cluster itself.