eMail David Southgate
david southgate writes for living.
david southgate: writing for living.
home published services musings about feedback contact

(Search on keyword)

(Or expand the menus below to browse by topic and headline)


Mailing List

Join my mailing list and I promise neither to spam you repeatedly nor sell your email address to anyone on the planet. Instead, I'll send you an email once in a while with a link to hot career and management advice— and maybe even a groovy travel destination. Why not join? [ list ]

IBM's Blue Gene Holds Promise for Research and Careers

published in February 2000

It's predicted to be so fast it'll have your proteins modeled in a petaflop

Imagine your NT box pumped up to two million times its current computing power, and you'll get a glimmer of the potential of Blue Gene, IBM's $100 million research venture. Announced in December 1999, Blue Gene promises to be 500 times more powerful than the world's fastest computers today. It'll handle an astronomical amount of computations (more than one quadrillion operations a second), making it 1,000 times faster than Deep Blue was when it beat world chess champion Garry Kasparov in 1997.

Blue Gene's first project - the modeling of human protein folds - might help medical researchers gain a new understanding of diseases and cures. But its practical application to real-world challenges probably won't stop there.

To learn more about the potential of this supercomputing challenge, spoke with Barry Robson, strategic advisor to the Computational Biology Center of IBM's T.J. Watson Research Center. Robson is himself a pioneer in bioinformatics, protein modeling, and computer-aided drug design.

Before joining IBM in 1998, Robson was an inventor in clinical therapeutics and scientific entrepreneur, working in pharmaceutical companies. He has published more than 175 papers, reviews, and patents, and has written a major textbook on protein engineering.

What is the promise of Blue Gene?

It's basically a simple processor in an extremely large array that will have the ability to process computations with no downtime, without resetting the computer if a chip fails. It routes the computation to a nearby processor and keeps the computation, without having to stop and go back a computer cycle to reset.

It's called Blue Gene because the first problem we're addressing with it is the process of protein folding simulation. By analyzing proteins, we'll be able to have better, faster, more accurate, and better drug design, for example.

We'll be able to analyze how diseases work at the molecular level. It's really about the physics of chemistry and being able to simulate and analyze protein in the computer while in some ways being able to reduce time to market for drugs and make the analysis more accurate.

You'd be able to load chemical data into it and manipulate it on a molecular level to see if you've created the right structure. It's a visualization problem with hundreds and millions of possibilities. That's why you need a supercomputer.

Are any other industries likely to benefit?

Other industries have visual needs that have a lot of data. The best way to understand it is to see a picture of it. These data all fall into the area of deep computing, a formal computing discipline. All sorts of industries [might benefit] - meteorologists, stock traders in futures, orange growers.

What opportunities (at IBM or research hubs) might exist for technologists wanting to work on the forefront of Blue Gene's development? At this point we don't have enough of a project to farm out, but we will have. There are people from industry and academia to help with this opportunity. We are going to see collaborators. We just don't have any assignments to give them at the moment.

What skill set does the team have right now?

The Blue Gene research team is, at the moment, composed of dozens of scientists at all different levels. So we have senior people and junior people. Basically, three disciplines are represented. At the moment, that consists of a research team with three working groups.

Computational biology

These people are Ph.D.s with 10 years in the industry who have worked in any of the life sciences, but also have significant experience in computer science - chemists, cellular biologists - people who've been involved in genome research.

Hardware scientists

The second working group is hardware scientists, people who are really focused on computer architecture. That encompasses chip design, computer communications (transmission technology within a computer), innovation in chip-to-chip, board-to-board, or unit-to-unit communication.

The computer has a huge amount of bandwidth with the unique design - the equivalent bandwidth of six billion ISDN lines. There are radical things like short cable runs between the units. So this computer represents a vision of stripping down a computer to its most basic elements in order to achieve a tremendous amount of computational power and speed. [Hardware scientists] do this in many ways, including putting DRAM on the chip.

Their degree is in computer science or physics, focusing on the hardware side of technology more than the engineering part of it.

Software group

The third working group is the software group. Right now they have the most senior people. They're working on radical designs for the software.

Right now, as planned, Blue Gene doesn't even have an operating system. It's what we call an "operating environment" because the nature of the problem is computational only. So it doesn't have to do other things like word processing. It just takes data and moves it to the next step incredibly fast.

So it has a very reduced instruction set. This is where part of the innovations of the entire project are to increase the computer's computational power and speed by constructing it in a way so it has just the bare minimum to accomplish its tasks.

So the software group is writing extremely tight code to enable the computer to run extremely fast and accurately. They work very closely with the architecture hardware group to understand how the system can exploit this.

How can Blue Gene potential help medical doctors in the future?

With genetic material you have to get it 100-percent right. It's what the medical field calls a kill-to-cure ratio. A drug has three states when in use. It either cures you (obviously the goal), it has no effect, or it kills you.

So when you're designing drugs you want to have the highest group of people helped by the drug. But it's also going to have potential side effects, which can range for minimal physical annoyances to death.

In the future, a computer will actually analyze a person's cellular biology, dissect your code, send it to a computer like Blue Gene, which analyzes it and helps doctors make decisions about highly effective, personalized decisions [about which drugs to prescribe].

This is the potential of the computer: to provide that kind of personal adaptation.

© 2002 - 2012 d.southgate   |   design: notlimited   |  

xhtml + css
david southgate
writing for living.