Horst Simon Steps Down As NERSC Director

Horst Simon, who has been director of DOE's National Energy Research Scientific Computing Center (NERSC) since early 1996, announced last month that he was stepping down in order to focus his energy on the two other positions he holds at Lawrence Berkeley National Laboratory. Once a new director for NERSC is hired, Simon will concentrate on his duties as Associate Laboratory Director for Computing Sciences and Computational Research Division (CRD) Director. With the search for a new NERSC leader officially under way, Simon took some time to talk about his decision and how he sees his future.

Question: For the past 10 years, you've served as the director of the NERSC Division at Berkeley Lab. Just over a month ago, you announced that you intend to step down from this position. Can you elaborate on your reasons?

Simon: For the past three years, I've officially held three jobs -- director of the NERSC and Computational Research divisions and Associate Laboratory Director (ALD) for Computing Sciences. During the past two years, the ALD position has become much more demanding. We are also initiating a new Computational Science and Engineering partnership with the UC Berkeley campus, while also planning to construct a new building to house NERSC and CRD.

I simply had to reduce my workload and get additional help to give Computing Sciences at LBNL adequate management strength. Relinquishing leadership of NERSC was the obvious choice as I've held the position for 10 years. The center is very well established, has secure funding, is being guided by a five-year strategic plan and we have very supportive program management from the Office of Advanced Scientific Computing Research in DOE's Office of Science.

The new director will have one or two years to fully learn the position, then can begin to look for new directions. It's always better to recruit in a situation of stability versus a crisis.

In short, I've been overcommitted and this is the right time to make a change.

Question: It seems that much of your identity is as the head of NERSC. What's been the reaction to your decision?

Simon: A number of the comments have expressed surprise. After that, people ask me what I'm going to do next. I'm going to stay at Berkeley Lab and build the new Computational Science and Engineering program at the Lab and on campus. And I'll be around to work with the next NERSC director. After I explain this, people say, "I see."

Question: You're only the fourth director of NERSC since the center was established 32 years ago. Looking at other large centers, there is a similar low turnover of directors. Does this surprise you?

Simon: At first, it did seem surprising. But thinking about it further, it was not really surprising. The job is a fascinating, all-engaging activity. What could be more exciting than leading a center with large national and international recognition? The technology evolves at a rapid pace and there are always interesting political ramifications. This is a field that clearly doesn't stay still. For a person who needs a daily dose of intellectual stimulation, this is one of the greatest jobs you can have.

Question: Given that such positions don't often become available, what is Berkeley Lab looking for in the new NERSC Director?

Simon: I think we've spelled it out pretty well in the position posting, which is now on the Web. I think the challenge for the new director will be to bring both operational and scientific excellence to the job. The person should have experience managing large-scale scientific projects or facilities, and have recognition and standing in the HPC community. We also want someone with a background in computer science or computational science -- a well-established track record.

Question: What attracted you to NERSC?

Simon: In the early 1990s, my dream job was to have an office in Berkeley, on the hill, running a computing center. I was working at SGI in 1995 when DOE made the decision to move NERSC from Lawrence Livermore to Lawrence Berkeley. When this happened, I could see the potential of NERSC in a more academic setting and was captivated. My reaction was "I gotta go there!" And I did.

Question: Of course, the NERSC Division Director position is only one of the three roles you currently have at LBNL. To pose a common job-interview question, where do you see yourself in five years?

Simon: I plan to still be sitting here. By then, we should have completed our new Computational Research and Theory building here at the Lab and have both NERSC and CRD moved into the facility. And we should have a large number of computational science programs built around our strengths in areas ranging from nanotechnology to astrophysics -- from the smallest scale to the largest. I also hope we will have a strong partnership program with campus in five years.

In the field of HPC, I hope we will have learned how to make effective use of highly parallel systems with hundreds of thousands of processors. In the 1990s, we first began to talk about "massively parallel" systems, but those systems typically had hundreds or a few thousand processors. The new machines coming on line now are truly massively parallel. We will have to make the transition to those systems, just as we will be transitioning to running applications on multicore chips.

Of course maintaining and growing our base research programs in mathematics and computer science, as well as our SciDAC projects will continue to require my attention. With a strong partner who will lead NERSC, I expect to be in a much better position to put more attention to nurture and promote computational research in the future.

Question: Anything else you'd like to add?

Simon: When heads of other centers have stepped down, they have remained close to HPC, which I also intend to do. There is a lot of work to be done in Berkeley, with computational researchers both on campus and at the Lab. With our new partnership, we hope to create a new environment to foster computational science, and NERSC will obviously be included.

One other area I intend to focus on is energy efficiency for HPC. I believe that increasing efficiency and lowering power consumption will be one of our big challenges during the next decade. I intend to devote a lot of my future efforts to finding solutions.