Patricia Teller

Patricia Teller talks about supercomputers with the same delight an adrenaline-addicted car aficionado might talk about the new 1,000-horsepower Mustang GT.

She’s the first director of research computing at the University of Texas at El Paso and she has a vision to build a sort of Tron-like cyber world in the circuits and wires that traverse UTEP.

Her cyber world would be complete with an infrastructure, economy and police force.

“Now the UT System is considering cyber infrastructure in a similar vein as the physical infrastructure such as buildings,” Teller says.

She has a little bit of funding and is starting small, but hopes to create a cyber infrastructure that is as fast as it is vast – a sort of research “cloud” where UTEP faculty members and researchers can store huge amounts of data and then mine it for answers.

And it could have the processing power to answer complex questions such as, “How are black holes created” or “What will the weather be tomorrow?”

Teller’s vision is to plug the UTEP cloud into the larger UT System cloud now under development.

Her office is quirky and is complete with a large, blue beanbag that has “Google” embroidered on it and exercise ball chairs.

“The students love those,” she says.

Although Teller does not have a “Tron” poster in her office, the figurines on the shelves reveal her love for the “Star Wars” and “Lord of the Rings” movies. A giant Yoda head hangs from her keychain and a Yoda figurine addresses office visitors when his hand is pressed.

Something like the plumber who has broken pipes at home, Teller says she prefers to spend her free time outdoors rather than on computers. She doesn’t have television but does love movies. She watches two a week from Netflix.

“My interests are all non-computing – horse riding, horse training. I like to read, I love wildlife and I like to hike,” she says.

Teller was born in the Bronx in New York City. Her father worked for a milk company, cleaning milk tanks and doing maintenance on milk trucks.

She was the first in her family to attend college, earning a bachelor’s degree, master’s degree and Ph.D. in Computer Science from New York University.

A faculty member at UTEP for 15 years now, Teller was named director of research computing in February. She has published more than 85 papers with students and colleagues and has been awarded nearly $7 million in research funds.

Wearing bright red cowboy boots, Teller sat down with El Paso Inc. and talked about petabytes, how cloud computing is transforming science, and which “Star Wars” movie is the best.


Q: Why this new position and why now?

In some ways, it’s simply because the stars are aligned, finally. I tried to bring in shared computing infrastructure earlier, in 2003 and again in 2005, as did others after that, but there wasn’t enough support and demand for computing infrastructure that could be used by others on campus.

Traditionally the scientific method was theory and experimentation, but computers have been changing that for some time now. When you use a computer for simulation and modeling, you can make the models much more realistic and complex than you can with mathematical models.

And with larger computers and memory, the models can grow in terms of realism and complexity. With this growth comes the potential to learn more about the problem and provide solutions.

Consider the complexity of predicting the weather, understanding how black holes are created, or testing materials under stress.

Q: Some might see what you are doing as simply building a giant hard drive and a super-fast processor for faculty and researchers at UTEP to share. But it sounds like the transformation is a bit more fundamental than that – that the “cloud” is actually changing the way some science is done.

Yes. First you had just the theory and experimentation. Now you have theory, experimentation and computation as well as something called data-intensive computing.

There are volumes of data out there and one can actually make scientific discoveries just by analyzing that data. We need the cyber infrastructure to support that if we are going to support UTEP researchers.

Q: Given that the UTEP cloud doesn’t really exist yet, what are researchers doing now?

It isn’t that faculty and researchers here don’t use clusters now on and off campus, but there is a proliferation of individual clusters around campus. University research is thought of as self-sufficient.

If you need a computer, you buy a computer with grant money or use a remote one and then you conduct your research.

For example, those who want to do large-scale computations as part of their research, write a grant, get a cluster and put it in their lab. But there are many problems with having clusters scattered around campus.

Q: What’s a cluster?

Basically, a set of connected computers with processors that work together to solve problems.

Q: How many of these clusters are there on campus?

We did do a preliminary inventory, but I’m not sure we have a valid number.

Q: What are some of the problems with the proliferation of clusters around campus?

There are many different reasons why that model is not a good model. Instead of doing science, researchers and engineers are administering and managing these systems. Also, the systems are vulnerable to security attacks if IT professionals are not involved in their administration and management.

Some labs don’t have sufficient power, cooling or space to support the systems. I know the room our cluster is in downstairs had to be retrofitted because the system was actually overheating – and that’s a small system.

In addition, it’s inefficient. When the researchers are not using their machine, it’s not being utilized. So here you’ve spent all this money and you’re not reaping all the benefits of that investment. Even from a green point of view there are benefits – a centralized system uses less power.

Q: So what is the solution?

That’s where the stars aligning comes in. The University of Texas System has started to invest money in cyber infrastructure.

Now the UT System is considering cyber infrastructure in a similar vein as they do physical infrastructure such as buildings, and it has invested some funds that would normally be used for buildings into cyber infrastructure I believe.

One project will interconnect all of the UT System institutions with the supercomputers at the Texas Advanced Computing Center, or TACC, in Austin, and the Arlington Data Centers. It’s connecting them with a 10-gigabit network, which is fast.

Q: Basically, you’re talking about a UT System cloud?

Yeah. TACC in Austin has supercomputers and it recently got funding for a new system called Stampede that will have a peak performance of 10 petaflops with 272 terabytes of memory and 14 petabytes of disk storage.

The UT System investment means researchers can get allocations on the machines at TACC. What they want to do is to surpass their peer institutions, attract superior faculty and students, and enhance the competitiveness of the UT System institutions for research funding. That fits well into UTEP’s goal to be a Tier 1 university.

Q: I’m guessing a petabyte is a good deal larger than a terabyte?

Way bigger. It has 18 zeros after the 1.

Q: When will all the UT System schools be linked up?

It’s close. There are a few outliers, including us, that aren’t connected. From what I understand, there is a connection in Downtown El Paso but the issue is what’s called “the last mile problem.”

To actually get that connection from Downtown to campus is the difficult part, but I understand it will be done by the end of the year. But the investment in networking, storage and advanced computing is just one part of stars becoming aligned.

We have a research data center opening here at UTEP. We also recently got an IBM Shared University Research grant and we are going to start building the UTEP research cyber infrastructure.

Q: Basically, then, you want to centralize computing at UTEP, building a UTEP cloud and integrating that into a larger UT System cloud.

Absolutely, for research computing.

Q: How much is the IBM grant?

It’s more than $100,000 in equipment and services, and that system will give us 84 cores, 336 gigabytes of RAM and 6 terabytes of data storage, and we have some funds that we are going to invest before June 1 that will increase the size of that.

I just got some quotes, and I am certainly hoping it will more than double the system size. What we would love to initially provide, although I don’t know if we will be able to, is 256 cores. That would give us a good foundation from which to build.

Q: What is the timeline?

The Research and Data Center should be open for business by May 1. I’m not sure we will have the Virtual Research Lab actually ready by May 1, but I am hoping we will have it ready by June 1.

Through the technology we will be able to give researchers access to the clusters whenever they need them.

Q: Something like renting a library book.

Yeah.

Q: But cloud computing raises its own share of security issues – all the data being stored in the same system and shared by multiple users.

The Texas Advanced Computing Center is at the forefront of this. They have professionals there who are experts in computing and storage security. This UT System cloud was actually pushed by the health institutions.

Digital medical records pose enormous privacy issues, so the systems have to have the level of security needed to store and access medical data.

Q: Will people be hired here to create and maintain the UTEP research cloud?

My hope is that this vision is adopted by many on campus and that it becomes successful. For example, North Carolina State has several thousand cores on their campus, and they have a team of two to three staff members who manage the system and help people use the system.

Eventually we would like to have students involved in the system’s administration as well as user support. The plan is to get funding so that we can have a few students to help those who might have some sort of problem using the system.

That, to me, would certainly help guarantee the success of this.

Q: Right now, what funding do you have to implement this?

We have the grant from IBM and some funding from the university. Getting more funding is going to be part of my job, as is working with faculty to buy shares of the system.

They will become partners who “own” part of the system; that’s part of the business model. I am working on a business case for this and a business plan.

Q: What is the business plan?

Basically, researchers who purchase shares of the centralized system will get a lot more than they would get by purchasing their own cluster. Let’s say a researcher gets a grant. What we are going to promote is investing in the central infrastructure instead of buying a cluster to put in their office or lab.

Because they will be adding to an already established infrastructure, we will be able to provide computational resources at a lower price than they could purchase them.

And software licenses can be shared, also bringing down the cost of ownership. We’re designing a website, brochures, posters and postcards that we are going to send out to promote this.

Q: Have researchers and faculty members showed interest?

We had a workshop in the summer and it was standing room only. Attendees were from psychology, fine arts, nursing, geology, engineering, computer science, biology, chemistry, physics, etc.

In terms of the data sharing and data discovery, it’s not only important to people who are generating a lot of data but it’s also important to people that have data collections.

The National Science Foundation and National Institutes of Health are actually requiring data management plans in grant proposals. For example, two of the chemistry researchers here at UTEP, they’re bringing up a cryo-electron microscope, which is key to their research. But the use of that microscope creates incredible amounts of digital data that have to be stored and analyzed.

That’s where the UT System network will come in real handy. They will be able to store that data in Austin and analyze it on the TACC supercomputers. The Institute for Oral History here has been collecting interviews for more than 30 years. They’ve had all their data on external drives and the drives started to fail.

Thank goodness they were also replicating the data and using good data management practices, but external hard drives do fail. So I had an undergraduate work with them to transfer all that data to the TACC data storage facility.

Q: Will local businesses be able to use the system?

Eventually I can see it as something that local businesses might invest in. Certainly in North Carolina they actually do that and TACC does that too – they have business partners. North Carolina also has community colleges and other colleges in the region who use their infrastructure, so that is also a possibility.

Q: Will students have access to the UTEP cloud?

Yes, for educational and research purposes.

Q: So the students can’t use the supercomputers for gaming?

No. I can see where that would be very attractive though. (Laughs)

Q: Students who graduate from the computer science program, what kind of jobs do they get?

I can talk about my own students. Some of the undergraduates will go into programming jobs and become testers and software developers.

Master’s and Ph.D. students, some of them become systems analysts or software analysts.

At the Ph.D. level, I have one student who is at the IBM Thomas J. Watson Research Center and another who is at the San Diego Supercomputing Center. I have master’s students at Microsoft, Intel, Cisco, Google and Apple.

Q: Is it hard for students to find jobs here in El Paso?

There is at least a reasonable percentage of our students who would like to stay in El Paso, and it is difficult for students to find employment in the area that utilizes their skills.

Q: If you had to play a futurist, what would you say the future of this technology is at UTEP?

If I were being a futurist, I might say that none of us need computational resources of our own. They will all be instantly available remotely – any configuration of hardware and software we want on demand.

Q: Now for the million-dollar question. Which “Star Wars" movie was the best?

I think I like the first. (Laughs) I really do. I like that one.


 

E-mail El Paso Inc. reporter Robert Gray at rsgray@elpasoinc.com or call (915) 534-4422 ext. 105.

0
0
0
0
0