May 2014

Where are you from, originally?

I was born in Munich, Germany, but spent most of my childhood and youth in Jülich, a small town in the Rhineland area—not too far away from the Belgian and Dutch border. Some colleagues and alumni from ICL have already been there—also because it is home of the Jülich Supercomputing Centre. The center, which is a part of Forschungszentrum Jülich, a major national lab, is located in a forest about a 30 minutes’ walk away from the street where I grew up. Further landmarks of the area include two strip mines, a sugar factory, plenty of farmland, mostly for sugar beets, and some historical fortifications. In addition, they recently constructed an experimental solar tower on the edge of town, which you can hardly look at during operation because its glare is so bright.

Can you summarize your educational background?

The high school I went to was located in an unusual place—within the walls of a centuries-old renaissance fortress in downtown Jülich, called the “citadel.” Every time I went to school, I had to cross a moat and pass through a tunnel like a knight, only I didn’t ride on horseback. The tunnel was dark because it was designed with a kink in it to prevent cannon fire from reaching the interior of the fortress.

Since I liked math but also had an affinity with language, I decided after graduation to study computer science. It was not actually that I felt a strong attraction to computer hardware or the desire to spend much of my time sitting in front of it. For my minor subject I chose physics, which enabled me later to pretend that I understood what all the simulation codes were about. I received both my master’s and my Ph.D. degree from RWTH Aachen University. My enthusiasm for supercomputing was sparked during my master’s thesis, as part of which I developed a performance-tool prototype that served both as inspiration and foundation for most of my research work to date.

How did you get introduced to ICL?

This was funny. I actually never applied directly for a position at ICL. A CV sent to another university in the U.S. that didn’t have a position for me was forwarded without my knowledge to Jack, who invited me to Knoxville for an interview. His email came as quite a surprise. The interview took the whole day, during which I had very stimulating face-to-face discussions with a number of future colleagues. After the technical part of the interview was over, Tracy Rafferty gave me a very nice tour of the campus. In the evening, Kenny Roche and Noel Black took me out to a restaurant and a few bars, including a martini jazz bar near Western Plaza and the legendary Cup A Joe. The interview was finished the following day in the early hours of the morning, after which I knew that working at ICL would be a rewarding experience and also a lot of fun.

What did you work on during your time at ICL?

Although I was technically a member of the PAPI team, Shirley Moore gave me the freedom to pursue my own R&D agenda, for which I am still very grateful. Essentially, I enjoyed the luxury of occupying a “bring-your-own-project” position. During my time at ICL, I therefore continued to work on KOJAK, a performance-analysis tool that helped HPC application developers to spot inefficiencies in their code. KOJAK had emerged from my Ph.D. thesis in collaboration with Bernd Mohr from the Jülich Supercomputing Centre. While I was in Knoxville, Bernd and I managed to move the project from a prototype to a release version—in addition to the many new features we added as a result of our research. That it was turned into a transatlantic partnership also made it easier to promote it in the U.S. Three ICL students, Fengguang Song, Nikhil Bhatia, and Farzona Pulatova, joined our efforts and helped us a great deal—especially with the interactive exploration of performance data. One year after my departure from Knoxville, Farzona visited my group in Germany as a summer intern.

What are some of your favorite memories from your time at ICL?

Because there are so many fond memories of my time at ICL, this is a tough question. One thing that made a vivid impression on me was the good sense of humor—sometimes even dry humor—I found among many people at ICL, and which I enjoyed so much. A magnificent scene that I still see in front of me was a lunch talk that Tracy Rafferty was supposed to announce but whose title she couldn’t remember. Instead, she stood in front of the audience with a straight face and said: “It’s about something scalable.”

Also, whenever I got stuck in a project, I went to Paul Peltz and Joe Thomas in the ICL help office, where we would crack some jokes, most of which I probably shouldn’t tell here. After that, I would return to my office in a more relaxed mood and untie the knot I was struggling with. In this sense, ICL help’s mission went far beyond purely technical questions and extended well into moral support.

Tell us where you are and what you’re doing now.

Since 2009, I have been head of the Laboratory for Parallel Programming at the German Research School for Simulation Sciences. The school is a joint venture of Forschungszentrum Jülich and RWTH Aachen University, committed to research and education in the applications and methods of HPC-based computer simulation. The lab is located on the university campus in Aachen, the city where I now live with my family. In addition, I am a professor at RWTH Aachen University, which gives me the opportunity to train students and, in particular, supervise thesis projects.

The lab’s mission is to develop methods, tools, and infrastructure to exploit massive parallelism on modern computer architectures. One of our key projects is the performance-analysis tool Scalasca, the successor to KOJAK, which we are developing jointly with the Jülich Supercomputing Centre. In Aachen, our focus is on automatic performance modeling with applications in requirements engineering for the co-design of exascale hardware and software. A major but relatively new topic that is separate from Scalasca is parallelism discovery, which covers techniques to guide the parallelization of sequential programs. Although it usually does not involve as many processors as our other projects, the number of applications that can benefit is quite large.

In what ways did working at ICL prepare you for what you do now, if at all?

Again, there are numerous things that I learned from working at ICL. What I would like to emphasize here is how much the success of a scientist depends on a powerful support infrastructure like the one provided at ICL, including but not limited to an effective business office, capable technical support, experienced communication specialists, and talented graphic designers. Another thing I learned to appreciate is a culture that values production software. As long as you are only creating prototypes to write articles in academic journals, only a few people, if any, will be able to build directly on your results, often reducing the practical impact of your work. Furthermore, you will hardly get any actual user feedback. Although not always flattering, such feedback usually contains very valuable guidance on where to invest your efforts.

Tell us something about yourself that might surprise some people.

As an eighteen-year-old, I had an uncommon hobby. I imitated voices—one voice to be precise. It was the voice of a very popular literary critic, who presented his book reviews on TV in an incredibly entertaining manner and whom I greatly admired. What started off as a party joke became quite serious when my classmates talked me into giving the farewell address at graduation, mimicking his voice and characteristic mannerisms. The subject I chose for my speech was a discussion of how literature was taught at our school. I then experienced intense stage fright until I tried the following advice I had found in a rhetoric manual. Before the speech, drink one glass of champagne—but not two.