Dr. Ing. Bernd Mohr
Deputy Division Head / SC17 General Chair
Jülich Supercomputing Centre
For the first time in the history of the Supercomputing Conference, an industry leader from outside the USA will serve as the conference’s general chair. The honor, slated for SC17, goes to Dr. Bernd Mohr, who has previous experience serving the SC Steering Committee, and served as executive consultant for ISC.
Since 1996 Mohr has been a senior scientist at the Jülich Research Center. He also serves as the Deputy Division Head of Application Support as well as a team lead for Programming Environments and Performance Optimization at the Jülich Supercomputing Centre (JSC). He is an active member in the International Exascale Software Project (IESP) and a leader in the European (EESI) and Jülich (EIC, ECL) exascale efforts.
HPCwire: Hi Bernd. Congratulations on being selected as an HPCwire 2015 Person to Watch. You’ve become the first person outside the USA to chair the Supercomputing conference. Can you talk about your experience with the conference, where it’s been, and where you see it going?
Thank you very much for this honor! My first SC was 1993 in Portland during my PostDoc days at the University of Oregon in Eugene. The conference was much smaller then, but it was an exciting time as it was when new parallel computer companies like Thinking Machines or KSR were founded almost every half year! It was also Intel’s 25th birthday and they threw a big party back then which I still vividly remember.
Later, in 1999, I was among a small team which was responsible for setting up and staffing the research exhibits booth of the Jülich Supercomputing Centre for a few years. I also gave 11 SC tutorials between 1999 and 2009. Also, I got involved in helping organize the conference as a research paper reviewer for SC 2003 and worked my way up serving various roles in the technical program committee. In 2010, I became the first European to serve on the SC Steering Committee.
For me, it was crucial to have this variety of volunteer experience so that I could better understand the numerous components and requirements of running a successful conference. Many people don’t realize how many volunteers are involved every year. Furthermore, each conference represents thousands of volunteer hours.
The conference changed quite a bit over the years. In 1997, the conference actually changed its official name from “Supercomputing” to “SC” to indicate that the conference was no longer just about supercomputing (= vector computing at that time) but also for cluster computing and MPPs (SC = Scalable Computing) and applications (SC = Scientific Computing).
So please stop calling it Supercomputing! 😉 Starting in 2005, the conference got its official tag line, “International Conference for High Performance Computing, Networking, Storage and Analysis” to make clear that there is more to HPC than just computing.
Unfortunately, the HPC community for many years still was obsessed by FLOPS which is also reflected in the SC technical paper topics: in the past two years only about 20% of the papers were about networking, storage or data analysis and visualization. So my goal for 2017 is to make topics like data-centric computing or extreme scale data analytics and visualization a much stronger and integrated part of SC’s program.
HPCwire: In this past year, the conference theme was “HPC Matters,” which seemed very relevant and well received. This year, the theme is “HPC Transforms.” Can you talk a little bit about the thought process going on behind these themes, and how you feel they are being received by the world at large?
HPC Matters is still an overarching theme for SC – in 2014 we made a three year commitment to support the concept with a goal of bringing HPC more mainstream recognition. You saw those efforts materialize in both the HPC Matters videos that we produced and the videos voluntarily submitted by organizations in our community. You also may remember the dynamic HPC Matters plenary at SC14 that featured SGI and NASA.
We will be doing similar activities this year. In fact, we will once again be actively looking for people to submit personal videos on why HPC Matters to them or their organization. Be sure to also keep an eye out for additional HPC Matters videos too! We also will be selecting a SC partner to give this year’s HPC Matters plenary. Look for more details in the coming months ahead.
HPC Transforms is the 2015 tag line and is based on similar principles as HPC Matters. It was developed by Jackie Kern the SC15 general chair and her team. This concept is focused more on continuing to showcase the people in the community who have a significant awareness of HPC and its importance to society. It is intended to tell the stories about how HPC is transforming science as well as lives for both society in general and the professionals within our industry.
HPCwire: Aside from the work you do with the Supercomputing conference, you have a day job Jülich Supercomputing Centre. Tell us a little bit about the work you’re doing there, and what we can expect to see from your group in 2015.
For many years, my research focus was on performance analysis tools for parallel programs. In 1993, as a PostDoc in Oregon, I invented and implemented the very first version of the now famous TAU toolkit. In Jülich, in the late ‘90s, my team and I started to work on automatic performance tools which not only go beyond just showing lots of collected data in the form of bar charts, histograms or timelines, but also track down known performance bottlenecks and their root cause and tell the user which and where our tools found them and how much they affect the program execution. Our current tool is called Scalasca (which stands for Scalable Analysis of Large Scale Programs) which is open source and freely downloadable at www.scalasca.org.
As the name suggests, we work very hard to make our tool work for large scale programs: just last year, we managed to measure and successfully analyze a real life physics simulation executing with one million MPI ranks on our BlueGene/Q in Jülich. Next year, we hope to provide fully integrated support for all major HPC programming models like MPI, SHMEM, OpenMP, Posix threads, OpenCL and CUDA or any combination of them.
HPCwire: On a more intimate level, what can you tell us about yourself – personal life, family, background, hobbies?
I am happily married for over 25 years. And I am only able to have such an exciting and enriching career because my wife has always supported me and takes care of things back home when I am traveling (which is often). We have a 22-year-old daughter and a 25-year-old son, both no longer living at home. There is still a lot of excitement in our house, as we own two beagles. When I am not working, I tend to be a couch potato: I can spend hours in front of the TV watching old and new episodes of my favorite show “The Big Bang Theory”.
When I was much younger, I was a very shy person. This is one of the reasons why I studied computer science, so I could sit in front of a computer and never would have to deal with people. Never could have I imagined that years later I would spent most of my time in meetings, workshops, and conferences with hundreds to thousands of people!
HPCwire: One last question – What can you share about yourself that you can share that you think your colleagues would be surprised to learn?
In order to earn money for my studies at the university, I worked every weekend as a disc jockey. These were the ‘80’s and it was the era of disco and pop music. “Gimme that night fever, night fever …”
I also like hiking and other outdoor activities. During my various trips to the U.S., I managed to visit 33 National Parks and many more National Monuments. At one point, I was close to having visited all National Parks west of Denver on continental U.S. but one (Big Bend in Texas). But the National Park service keeps “upgrading” National Monuments to Parks, so I do not know whether I’ll ever reach that goal.