Supercomputer project a super IT stunt

Stunt IT can be applied to all IT, says Frank Hayes

At Purdue University recently, the IT staff built a supercomputer. A world-class supercomputer. Out of PCs. Using just local IT talent. (OK, a team from Indiana University also helped). And it was done in less than a day.

A stunt? Certainly. A real IT project? Yes, just as certainly.

In fact, it was the epitome of a real IT project. It was funded by the departments that will use the machine. It was designed so the right technology can solve problems, and planned so that it would come together cleanly and effectively. And it went operational on budget and ahead of schedule.

How many real IT projects can make all those claims?

Of course, it was also a stunt. The new supercomputer was given a catchy name, "Steele" (after retired Purdue computer centre director John Steele). The project was announced with a YouTube video made to look like a movie preview. More than 800 off-the-shelf commercial PCs were delivered, unboxed and rack-mounted all at once, just to show it could be done.

And it was assembled and working in one day. That wasn't just because it made a better stunt. The unniversity's three existing scientific-computing clusters had been taken offline five days earlier, and academic researchers needed the hardware up and running fast.

When was the last time you did something like that? Something splashy and flashy, something to make users say, "Wow, this is really great stuff"?

It's a good idea, doing a stunt now and then to show off. It's good for morale and for IT's reputation in your company.

But you wouldn't want to do it with every rollout. Or even with every major rollout.

See, a stunt has to work. It's intentionally high-profile. The point is to have everyone watching as you look good doing what you do. You really can't afford to fail.

That means the flashy stunt rollout has to be planned with extraordinary care and executed with extraordinary discipline. Everyone has to know exactly what's supposed to happen. Each IT person — there were more than 200 involved in the Purdue rollout — has to know his task and be prepared for whatever might go awry.

And as with any good magic trick, everything has to be prepared in advance to minimise failure. That requires lots of advance testing, well-practiced hardware configuration and bulletproof software installation. Everything that can be done before zero hour should be done.

So, no half-baked plans, no seat-of-the-pants navigation, no individuals doing it their own way — just a smoothly executed delivery that makes users happy and makes your IT staffers look brilliant.

And there's one other advantage: Stunt IT is fun. For most of the IT staffers involved, it's a burst of hard work followed by an enormous feeling of accomplishment and plenty of admiration from the rest of the organisation.

There's really just one downside. No, not the chance of failure — with the right preparation, you can do this.

The downside is that once users see your IT shop deliver a crisp, disciplined, well-planned and well-prepared rollout, they'll expect that level of quality from IT all the time.

Then again, if you have the skills and discipline necessary for stunt IT, you should be able to apply them to all your other projects. They might not be flashy and splashy, but they'd be the sharp, effective, real IT work your business needs.

That really would be some stunt.

Join the newsletter!

Error: Please check your email address.

Tags managementPurdue UniversitySupercomputer

Show Comments

Market Place

[]