One of the most famous experiments conducted at the HNL was a double-blind taste test pitting Pepsi against Coca-Cola. Neither the researchers nor the subjects knew beforehand which soda they would taste or in what order. "They were lying in the scanner, with tubes in their mouth, and they'd get a squirt of Pepsi or Coke," King says. "Hundreds of people went through this study. What we found was, when you don't know what you're getting, people prefer Pepsi. When you do know, people prefer Coke by a huge margin." And the interesting part of this equation is that the researchers didn't have to ask the subjects which drink they preferred. Instead, they could tell by looking at the strength of the response in the brain's "reward center," as measured by the fMRI scanner.
King wrote an open source application called NEMO that makes it possible for researchers to conduct collaborative multi-subject experiments from several different locations at the same time. Each scanner is set to start collecting data at precisely the same moment, as the subjects inside the scanners play a computer game together or engage in some other kind of cooperative activity. "Nobody has ever done anything like this before," King says. But no one really knows what to do with all the data. "Basically, we found out that it can be done; that doesn't get you very far. The analysis becomes ultra-complicated with four brains. Nobody knows how to even approach it." Still, curious researchers continue to write scripts for NEMO to process. "We do a lot of two-person scanning, with someone at Baylor and someone in say, China or Hong Kong University." One of the NEMO experiments is studying the cultural differences between participants in trust games. "We've developed this nice baseline as to how the brain reacts in this given situation," King says. "Now we can start doing more interesting things like scanning autistic children or Parkinson's patients or borderline personalities to see if there are cognitive deficiencies in an area like trust, which is extremely important for survival. We're trying to find people where that is broken."
The lab recently added a Computational Psychiatry Unit (CPU) that will house more medical research personnel and double the size of the network. With the CPU, faculty members will have their own lab area making use of the HNL scanners. King says he is determined to stay "out in front" of the expected onslaught of bandwidth usage and network complexity.
In anticipation of the increased load, King deployed the enterprise version of Hyperic HQ open source network monitoring software. "We're not ultra-critical to be up 24/7, but people do work all the time," King says. "The students have no choice but to put in 15 hours a day, and it's important to me personally that [the network] is available as much as possible.
"My infrastructure is about to explode. I need to be able to respond, and I'm still the only guy that's running everything. I need to be out in front of that, I need to know where the issue is and what's causing it. The academic types want something when they want it. They have some pretty good-sized egos," he says. "I don't have a lot of time to spend troubleshooting."
King looked at OpenNMS and Nagios before choosing Hyperic. "I really didn't look at anything I knew I couldn't afford, or that was overkill for what I was doing. I had been using OpenNMS and I liked it for what it did, but it only gives you 'this is either up or down, this port is either open or not.'"
Nagios was more complicated than King preferred. "One of the main complaints about open source in general is that you have to install a lot of different pieces, and while none of this is particularly difficult, it's still sort of a pain."
King says he heard about Hyperic from a Slashdot post, downloaded the free version, and installed the software in one step that took him no more than 15 minutes. "I thought, 'How easy is this?'"
King didn't have to buy the Enterprise Extensions for Hyperic -- he just wanted to. "We use so much open source, I can't add up the value of it. When you see a good project you want to encourage it. I felt a community responsibility to support Hyperic. I believe in the project." With an educational institution discount, King says he ended up paying about $15,000 for the deployment of the enterprise version. "It was a number that sort of gives you pause -- but at the same time, when you start thinking about the TCO, it would take me two months to write the functions [included with the enterprise version]. All of a sudden it makes a lot of sense."
King says he doesn't feel compelled to use only open source, and other IT managers shouldn't either. "You need to use whatever tool works best for you. In our case, it makes sense for me to use Windows as the authentication domain, for example. That's easy to do. At the same time, the open source stuff works awesome for what we need it to.
"You get so much use out of the open source pieces in general. Apache, for a quick deployment to get something up, it's cake. It takes five minutes. I'm looking for whatever's easiest and most efficient for me, and that will present the best experience to the user."
King says he never buys support contracts -- hence the CentOS servers and clusters instead of Red Hat. "I never use it," he says. "I get a kick out of figuring out what's wrong. If I don't have the answer to a problem, the community is a very powerful tool. That's all the support contract I need."