MaSSSoftech

It wasn't long ago that universities were considered the poor cousins of big corporations when it came to information technology. Students learned how to program, got jobs at big companies and breathed a sigh of relief at just how quickly and efficiently technology was modernized and upgraded.

Fast forward a few years. Data center managers are now struggling to clean up layers of old servers and applications, and become more responsive to their business units. Universities, meanwhile, have rekindled their experimentation with technology, using it in ways that would likely make their corporate counterparts shudder.
Forbes caught up with Gerry McCartney, CIO at Purdue University, to talk about the changes under way in academia.

Forbes: What does the CIO job involve at a university?

Gerry McCartney: In theory, it's anything that's a production service. We support technology in the classroom, run administrative systems--administrative support and payroll--and we have all the teaching systems. We also support research computing.

Do universities make use of cloud computing like many companies?
Absolutely. We've been making very aggressive use of cycle sweeping, which is an earlier form of cloud computing, and grid computing. We have a virtual grid of 28,000 CPUs (central processing units), which draws waste cycles from five other institutions as well as Purdue. A repackaged version of that is what we think of as commercial cloud computing. With that you're not using specific machines. You're using a service that makes those cycles available to you.

How do you allocate that compute power?

If you think about payroll, you may run that every week or two, but you don't need to run it all the time. You can predict your demand and model that. What researchers want is as much capacity as possible. They'll consume whatever is available, so our research machines run at 95% capacity all the time. There is such demand that we start processing research jobs even as we're building new machines. But with research computing there's less demand to do it today. Payroll is a time-critical service. Research is not. So if I can't offer cycles on Friday, I might be able to offer them on Sunday, and for most jobs that's acceptable.

That's way above the highest utilization inside corporations.
If they have research departments, it's probably that high. But there's a lot of technology that's just sitting around waiting inside companies. What we do is stack up the demand in a queue and have them manage that queue. We tell researchers to buy the compute nodes, and we pay for everything else--the inter-networking fabric, storage and backup--but when they're not using their nodes, we want to be able to use them. If you bought 16 nodes on a big machine, you get to use them anytime. But you also can grab 32 nodes on a neighbor's machine. And when you're not using your 16 nodes, they're available to others.

Does this work on a global basis for companies, as well?

It can. The problem is if you're moving around large blocks of data. That can cause latency problems. If you have to move terabytes of data to Hong Kong, it can take hours and hours to get it there. Or worse, you send a couple gigabytes, and when you're done you have terabytes of output. That's hard to get back. The compute capacity is still somewhat ahead of the networking capacity. You still want storage near your compute environment.

Are there other technologies in use in universities that might apply in the corporate world?

At the retail end, we don't understand how to use social-networking technologies like Facebook, MySpace or Wikipedia in a productive way. They're still toys, and corporations have danced around the edges. But a lot of their employees are using these things, whether they like it or not. We've been experimenting with ways to take advantage of these new technologies and engage students in a whole new way of learning. The students are actually driving us in that direction.

It's a different baseline, right?

Yes. Try explaining to a student today what a phone booth is. They don't get it. We're going to see the same in education soon. The old one-to-many model, which is the traditional educational or broadcasting model of a talking head with no ability to interact, is being transformed. The best analogy is television news. You've still got the talking head in the middle of the screen, but you've got all this other action on the screen. You've got scroll bars on the side, you've got a ticker on the bottom. People can watch three or four different things simultaneously. The talking head is delivering information at one speed, and the banners and scrolling information at the bottom are being delivered at a different speed. Your brain can actually absorb that quite easily once you get used to it.

How does that apply in education?

I can be listening to a professor and at the same time have the equivalent of a Twitter line open to the teaching assistant where I can ask him a question about what the professor is talking about. The teaching assistant can answer the question in real time. The teaching assistants are being used to supplement education, and the faculty is watching these background conversations to see what's going on. If a whole group of students is asking a question about what is an internal rate of return, for example, the professor can stop and address that and then move on. We also have an algorithmic filter that removes all the noise on Twitter like, "Me, too" or, "I agree with that."

Are grades going up because of this?

We're not sure yet, but the old idea of, "Look left, look right and one of you won't be here next year," is an incredibly wasteful process at the institutional level and at the individual level. No one is thanking us if their kid lasts two years and then drops out. This allows us to say, "If we admit you and you're willing to work, you're going to graduate with a degree in a timely way." We're not going to try to make it hard for you. We're going to keep you focused and on task. We also use technology that allows us to detect in the first two weeks, based on the way you interact with online course material, whether you're at risk or not. We run intervention. The professor can say, "Come visit me after class, go visit the teaching assistant, or do this assignment over again." Students love this stuff.

When did you begin using this technology?

Just about two years ago. We have 8,000 students using it this semester. So we can't tell the ongoing impact yet. But we're watching it very closely.