The Bloor Perspective: IBM and grid computing

In this snapshot look at a recent key development, Robin Bloor and his team consider which the concept of 'grid computing' and an IBM initiative

What happens if you network tens of thousands of PCs together and share the resource? In theory you could assemble many teraflops of computing power and just might be able to use it productively.

This is what AFM (the French Myopathy Association) and IBM are hoping to do. A French Telethon which takes place on December 7-8 is going to invite PC users in France, or anywhere else in the world, to make the processing power of their PCs available for use on the "Decryption" project. The project involves mapping over 500,000 proteins in an attempt to understand them and their relationship to genetic diseases.

The comparison and classification of proteins is a compute-heavy task. AFM and IBM will be employing a software application, developed by Geomining and based on the Smith-Waterman algorithm (whatever that is), which will work in a parallel manner, sharing the workload between a server and, hopefully, a vast network of PCs.

If the application were run on a single PC, it is estimated that it would take about 1,170 years to complete. Even on an appropriately configured cluster of supercomputers it would probably take a decade. With the help of the general public, AFM and IBM are hoping to complete the project in months.

At France's annual muscular dystrophy telethon in December an appeal will be made for the home PC users to make their PCs available to the project in order to create a massive virtual supercomputer with a calculation power of up to 40 teraflops (40 trillion calculations per second).

In February 2002 participants will get an email with directions on how to download the application. Once downloaded and installed, the PCs spare processing power and bandwidth will be used to exchange data when the PC is online and to do calculations whether online or offline.

There should be no visible impact on the PC as the application is designed only to utilise unused capacity. Researchers are hoping to assemble about 100,000 contributing PCs and get about 100 hours of time from each.

This is not the first time that a big virtual network has been attempted - see SETI and Stanford University's Folding@Home. However, this project is slightly different in that IBM is hoping to learn a great deal about the possibilities of 'grid computing' from it.

The grid is the possible future of the internet. In effect the idea is of an information utility that spans the world and involves participants in renting out resources to users, hopefully at profitable but inexpensive rates. However, managing and sharing the resources of a vast network is not a trivial task. Software needs to be built to allow this to happen in a controlled and robust manner. The best way to solve this problem is to assemble a big grid and see what problems emerge.