INDUSTRY

Interview with Wolfgang Nagel, Director of the Center for Information Services and High Performance Computing (ZIH) and Professor of Computer Architecture at the Dresden University of Technology, Germany. In 1997, Wolfgang Nagel left Germany’s famed Juelich Research Center and took a position at Dresden University of Technology. In Dresden, located in the former eastern part of Germany, he has led the development of the Center for Information Services and High Performance Computing (ZIH), which now comprises a staff of more than 70 researchers, students and guests. The center was recently renamed to better reflect the integration of the HPC center (ZHR) and the university’s computing services center (URZ). There are also other changes in the works, including the installation of new computing systems and a next-generation release of VAMPIR, the flexible tool for tracing performance issues on parallel systems which Nagel and his group has been working on for nearly 10 years. The center conducts computational science research in biology, materials science and turbulence. Additionally, one team is involved in benchmarking, an important area as the ZIH has just procured a major new system. Members of the ZIH staff are also working on HPC performance analysis, modelling and optimization, furthering Nagel’s development of VAMPIR. This interview first appeared in Primeur. The publication recently caught up with Nagel for an interview about the current state of the ZIH, its future plans and HPC-related activities in Dresden. Primeur: Your center recently co-hosted the European Conference for Mathematical and Theoretical Biology (ECMTB05), you have been selected to host EuroPar 2006 and at the ISC2005 conference in Heidelberg, conference organizers announced that ISC2006 will also be held in Dresden. What’s going on here? Prof. Nagel: Well, from our perspective this is additional recognition of something we have realized for several years. The German state of Saxony has been actively developing a strong economic base with high-tech manufacturing, biotechnology and other leading edge activities. The Dresden University of Technology is central to many of these activities, and with the deployment of our new HPC system, we believe that this growth will continue, if not accelerate. All of these pieces are now coming together. Primeur: The procurement of this new system has been under way for some time. Where do things stand now? Prof. Nagel: Things are great! We have now gotten all the necessary approvals and have awarded a contract for the system. Installation is scheduled to start in September and conclude in 2006. When the new 2,000+ processor system is ready, we will be a very capable mid-size center, by design. Our center will focus on data-intensive computing. We will have a very balanced system, but it will be composed of heterogeneous components in all areas. It will be built around a SAN infrastructure together with a petabyte archival system. I strongly believe in the heterogeneity of applications. I don’t think one hammer can be used to nail down all problems. I think our approach is part of the natural process of addressing the data problem. Part of the new system will be a PC farm based on the AMD Opteron processor , which will allow us to serve many people with different resource requirements. For example, there are many users who would benefit from using 2, 4 or even 8 CPUs, and this farm will provide that. It won’t be a cluster for parallel jobs, but will have excellent I/O for jobs running on just a few or even one processor – applications such as BLAST. The HPC component – an SGI Tollhouse system based on Intel Itanium processors - will provide more than 4 terabytes of global address space, especially interesting for scientists in the field of biology. We think the overall system will be an innovative tool to help them investigate their data on a much higher scale – from 32 gigabytes to 4 terabtyes, which can be used for interactive investigations of data. In such a system, the memory bandwidth is the critical part, both from memory to the CPUs and from the SAN to the CPUs. This bandwidth will be at a very high level in our system. Both biotechnology and bioinformatics are growing fields and our center will support them. In the future we will be supporting other research institutions and universities in Saxony – not just the Dresden University of Technology. I’m very excited by the prospects before us. Primeur: The ZIH in Dresden may not have the same name recognition as other centers in Germany, but that sounds like it may be changing. Your staff seems to be quite active in a number of key research areas. How would you describe the current state of ZIH? Prof. Nagel: After eight years of development in the former organization ZHR, we have established a couple of interesting research areas in which we are focusing and making significant contributions. Within the HPC community, I think there is growing recognition of what we are doing here. For example, the area of performance analysis is one that the ZIH should be – and is – addressing. I’m very happy with the continued development of VAMPIR, which is now successful on a very scalable level with a major commercial company. It’s in the market and in use, supporting major research centers worldwide. VAMPIR will be 10 years old in January and we have something special planned. We will be releasing VAMPIR NG, for next generation. We expect to have a prototype version generally available by October, but a supported version available to purchase by Jan. 1, 2006. For the past couple of years, we have also been looking to the field of biology. We now have a well-rounded group working on such issues as self-organizing systems, and seeing how what occurs in nature can be applied to systems constructed by humans. The State of Saxony is also making a major investment in biological research, supporting many activities in this field. This is one of the reasons why our center is also moving in this direction. It’s important for us to have our own staff who speak the same research language. Our Innovative Methods in Computing group, which is led by Andreas Deutsch, can serve as an interface between ZIH and the science community. It’s important for us to bridge the gap between computer scientists and researchers because, for the most part, we are likely to be engineers than science people. The third area where we are contributing is the development of a reasonable architecture. We can now support our end users on our machines. Our expertise allows us to help optimize their applications and we have the necessary experience in administering midsize systems. All computing aspects are now well supported, so our users can now focus on getting results, rather than waste their time working on computer problems. In short, I think the ZIH now has strong efforts in performance optimization, modelling and analysis; in supporting biological research; and in providing a good level of service to our users. Primeur: That leads to the next question. Biology clearly represents a major thrust area for both ZIH and the German research community. Where do you see your center making contributions? Nagel: One of the major requests from biologists is support for data-intensive computing. Just last week I received another call from a computational biologist who wants to meet with us. Computational biologists are probing the limits of computing and data storage. Many of those biologists started out on their desktop computers, but now see a need for a bigger infrastructure with larger computers. They want to get more information out of their existing data, and they want it to be easier to handle this complex situation. This is the challenge for bioinformatics over the coming years. This is a new approach for us and we are working to especially support this growing community. Developing the infrastructure to do this has been a primary focus of ZIH over the past two years or so. We designed our new system with these requirements in mind. Primeur: Coming from one of Germany’s best known research institutions to a relatively lesser-known computing center must have required some adjustment on your part. What’s been your experience in Dresden? Nagel: To be honest, it’s been a lot of fun to set up a new center. We need the dedication of all the participants all the time, to be ready to focus and to be fast on our feet. It’s really fantastic how things have developed over the past eight years. We have gotten new funding and we’ve been able to grow. We started pretty much from scratch and are now recognized as a contributor in different areas, from algorithm development to support for performance tools. Much of our research funding now comes also from outside the Dresden University of Technology and I think this represents a real recognition of our capabilities. What we want to achieve here is to be recognized as an excellent midsized center, excelling in data-intensive computing. We want to be an attractive point for someone with problems to solve in this area, not just within that State of Saxony, but beyond. What was envisioned for here was what my research department at Juelich was focused on. Here there was an opportunity to shape a new center and to make it responsive to our users’ needs. In terms of adjusting, it’s been a very positive experience here with a lot of support form the university management. It’s been enjoyable from the beginning. We’ve had eight years of winning, of gaining expertise and experience – and we’re doing nice work. Primeur: In other conversations, you have mentioned how Dresden is an ideal location for advancing science and computing. Besides the fact that the coffee filter was invented here, and that most research today seems caffeine-driven, can you elaborate on this? Nagel: Dresden is a very interesting town and is developing very well, especially compared to other cities. We’re very lucky to have three Max Planck Institutes, 10 Fraunhofer Institutes and several Leibniz Institutes in the region. The Technical University celebrated its 175th anniversary two years ago, is one of the oldest technical universities in Germany and offers excellent engineering capabilities. The number of students has grown from 24,000 in 1997 to about 34,000 today. In principle, it’s a very attractive point for students to take their first steps in research. The university also has a very successful computer science department, which has grown to about 700-800 incoming new students each year. This is one of the largest computer science departments in Germany, larger that TU Munich, for example. And as I mentioned, we have a good infrastructure here and as part of that infrastructure, we can use the infrastructure to collaborate with others. For example, the Bio Initiative of Saxony, as I mentioned, represents a significant investment – and an opportunity for us. With the Max Planck Institute for Molecular Biology and Genetics, we have an opportunity to work together on new and innovative solutions to real problems. As a computing resource for the entire State of Saxony, we will also computationally support researchers in Leipzig and Chemnitz. While Dresden, and other cities in Saxony, are known for their history of engineering breakthroughs, not everyone knows that the idea of a personal computer, a computer that would sit at your workplace, was actually developed here at the Dresden University of Technology. Back in 1948, in response to reports of ENIAC in the USA, Joachim Lehmann of the university began to research and develop a programmable computer. But, because of the lasting damage and shortages from the war, Lehmann decided to focus on building a smaller computer than ENIAC. The D1 (or Dresden-1) was a large system, but the subsequent version known as the D4 was about the size of a large TV. Six of these were built in 1956, and 3,000 examples of a variation known as the C 8205 were built in the 1960s, so I think it would qualify as the first personal computer. The university was also home to a BESM-6, a Soviet-built supercomputer, in the 1970s and 80s, so computing is hardly a new concept here. Primeur: Your center has been selected to host the EuroPar2006 conference, and your staff has hosted the European Conference on Mathematical and Theoretical Biology - ECMTB05 with about 800 participants. The ZIH appears to be making a real effort to attract high-profile European meetings. Prof. Nagel: As I mentioned, both our center and the city of Dresden, as well as the state of Saxony, are building on a number of strong foundations and emerging as key research components of a growing research community. We think that by hosting these conferences, we can continue developing relationships with others, while also sharing our expertise. We were also excited to learn that the International Supercomputer Conference will be held in Dresden in 2006 after 20 years in the Mannheim-Heidelberg area. This will be a good opportunity for us to increase our visibility in the HPC community. Primeur: Where do you see the ZIH in 2007, as it marks its 10-year anniversary? Nagel: After the installation of our new system is completed in 2006, we will have a powerful and well-balanced infrastructure for achieving excellent results with applications. Our goal is to help our users make significant progress in their research. By the same measure, I’m sure we will have addressed new questions in our own fields of performance analysis and computer architecture – and this will help us prepare for the next 10 years. However, our users can only generate good results if they get excellent service. We need to be able to see what their needs are, and then meet them. But what I would like most is for us to be recognized as a high performance computing center for solving interesting research problems. By continuing our partnership with the university leadership, I would hope we would be in a position to help the university attract top researchers. We would then provide the systems and support needed to attract researchers with new applications in engineering and the natural sciences. What I hope is that scientists who are looking for a good environment to do their research, an environment that includes a strong computing infrastructure, a good living environment, a center of culture, then they would see Dresden as an ideal environment for their work. Supercomputing Online wishes to thank Jon Bashor (LBNL Computing Sciences Communications Manager) and Ad Emmen (editor of Primeur) for their help with this story. For more information about ZIH, go to its Web site.