FEATURE

The man who built ISC

John Murphy meets Hans Meuer, who has been involved with the International Supercomputing Conference in Germany since it began.

There are few better known names in the high performance and supercomputing world than Hans Meuer. He is the general chair for the International Supercomputing Conference held in Germany – and has been every year since it began in 1986. He built it up from a small meeting of German supercomputing enthusiasts in Mannheim into the premier conference outside the US in its field.

Meuer has been involved in the field since it began in the 1960s and he knows everyone in it. As well as the conference, he started the Top500 league table of the most powerful computer installations in the world, making him a key player in the global competition between the ‘big iron’ sites.

His conferences are renowned for their family atmosphere and the tone is clearly set by his big, warm and outgoing personality. He knows the top speakers personally and they find it difficult to refuse his invitations to speak. Everyone they need to impress will be there, so not only does it attract the up and coming scientists, but the top vendors have to be there too.

Jack Dongarra, Distinguished Professor of Computer Science at the University of Tennessee and co-founder of the Top500 list, says: ‘I have been going to Hans’s meeting since the early days; I go about every other year now. I can remember doing a talk and I was the only person there who was not German, but since then it has become international. It has a very nice feel to it and it complements the big meeting we have in November in the US. ‘He is a very engaging guy; upbeat and happy. He knows the European high performance computing crowd and most of the US people. He certainly has the respect of people within the community. Every meeting is like a homecoming, with him greeting all the people he knows. The meeting really is him, he pulls it together and it’s all about his drive. The meeting reflects his personality in that it’s a friendly, family-type event. It’s a very relaxed setting and you get a chance to talk to key people.

‘His meetings have given people a chance to see where everyone else is, and discuss issues for which there normally isn’t time. Sometimes vendors make announcements there, and there is usually a technical content. Hans organises good speakers who have something relevant to say.’

Dongarra says that his knowledge of Meuer was from the Top500 project and the conference, but that his own personal scientific work has, in many ways, been overshadowed by his success as a conference organiser. But he says Meuer’s contribution to the field of high performance computing and to the economic success of his country is a legacy to which anyone would aspire.

He says: ‘He came into his own at the height of supercomputing in the 80s and rallied the German community first in terms of understanding that supercomputing was important strategically, giving one competitive advantage and the German economy has benefitted from that. Germany has the largest collection of supercomputers in Europe and Hans has contributed to the understanding of their use considerably.’

Meuer was born in 1936 in Laubach, a small farming town about 60km north of Frankfurt. His father was a school inspector and his mother was a home-maker and they lived together with Meuer’s twin sister and an older brother. He showed an early interest in mathematics at school, but his main hobby was playing chess. He joined a chess club and became master of the club within two years. When he went to college he gave up chess, because he knew so many people who had not finished their studies after spending too much time playing chess. He took it up again at the age of 40 when his own children started to play. Meuer’s father had wanted him to study to become a teacher and so he went to University in Giessen, spending one year in Marburg and a semester in Vienna as part of his studies. But when he graduated in 1962 he decided to defy his father’s wishes.

Meuer says: ‘I didn’t do a Master’s degree in mathematics, which was my wish; my father was in favour of my taking the course that allowed me to become a teacher. But when I finished I decided not to be a teacher, against my father’s wishes. I had some interest from a professor at the Jülich Nuclear Research Centre, who had gone there from Giessen, and he asked me to come there with him.’ His course required him to take another subject, so as well as doing a dissertation in physics and mathematics, he also had to do a dissertation in politics and, as he was always interested in history, he decided to study the resistance against the Nazi regime. He says: ‘I was very happy to do that and I was very happy that I did not go on to become a teacher.’

Meuer’s conferences are ‘renowned for their family atmosphere and the tone is clearly set by his big, warm and outgoing personality’.

Meuer started using computers as soon as he joined the Jülich Centre in 1962. At first they had an IBM 1620 mainframe and established a computing centre from scratch. A few years later it got an IBM 360/75, which at the time was an extremely powerful machine, and only the second in Europe.

He says: ‘I cannot claim that these were supercomputers, but they were extremely powerful machines that were used for mathematical calculations.’ While working at Jülich he was invited to study for a PhD and, despite now being a group leader and having a young family, he managed to gain his degree in 1972. This was the last time he got involved in mathematics because, in 1974, he moved to the University of Mannheim, where he worked entirely in computing.

Mannheim was not a technical university and specialised mostly in business. It did have a department of mathematics, but it had not been involved in large scale computing. Meuer says that while at Jülich he had become interested in the high performance computers of the day.

He says: ‘Everything was just starting at the time and high performance computing was more or less my hobby and had nothing to do with the everyday work of Mannheim. I did eventually persuade them to buy a Kendall Square machine; the company had gone bankrupt and I got a very good price from them. I did give lectures in parallel and high performance computers and, of course, I had access to Cray machines at other institutions such as the German Aerospace Centre (DLR). I was always a user of these machines, but we did not operate such machines for a long time.’ Meuer decided that it would be a good idea to hold seminars with some of the vendors. The first was arranged at Mannheim, with Burton Smith as guest speaker and a second was arranged at DLR with Hitachi.

Meuer says: ‘In 1986 we decided to start the Mannheim Supercomputer Seminar; we did not know how long we would do that for, but the first time we had 81 participants. At the time the only supercomputer in Germany was the Cray 1 at DLR. Originally it was for German participants, but people started coming from all over Europe and the US. This year we hope to have 1,300 participants. His meeting has grown into the International Supercomputing Conference. Despite the fact that it has started to move location around Germany, it is still known in the field as the Mannheim Conference. Meuer says: ‘We started in the university and eventually we moved to a conference centre. Later, we needed somewhere bigger, so we went to Heidelberg for five years. We decided that we could not run the exhibition there so we moved to Dresden for the last three years and next year we are going to Hamburg. ‘We have many participants who come every year, so we think they might be happy that after three years in one place they are happy to move to another city.

‘I think it is very important that people from this field get a focus and chance to meet each other and find out about what is happening. The advantage of our conference is that we are not as big as the US conferences and so you have a chance to get in contact with the speakers and talk with them at lunchtime. The US conferences are too big. We have about 250 to 300 coming to our conference from the US, because there are many advantages for our conference.

‘I think our speakers are usually more wellknown than in the US. Of course, they might get Bill Gates as the keynote speaker, and we cannot get Bill Gates, but we have speakers like Horst Simon and Jack Dongarra, who love to come to our conference and people like to get the chance to speak to them. ‘Most of our speakers are hand-picked and we like to choose hot topics, but there are also chances for younger scientists to submit papers.’

One of the most famous outcomes of the Mannheim Conference was the Top500, a league table for the most powerful supercomputers. There was always a discussion at the conference about who had the most powerful computer. In the early days Meuer used to quote statistics given by the various vendors together with information supplied by friends. By 1992 Meuer had compiled a list of 530 supercomputers in the world. He realised that changes in the technology were making the list more difficult to compile. Firstly there was the emergence of ‘minisupercomputers’, and there was the start of the drift away from vector processing. There had to be an independent measure of the actual power of a supercomputer for the league table to be of any use.

Meuer worked with Eric Strohmaier, then at Mannheim but later at Lawrence Berkeley, on compiling a new league table and Jack Dongarra suggested using his LINPAC software package to create a benchmark, because it was very easy to get performance statistics out of it. Horst Simon from Lawrence Berkeley Lab was involved in the early stages and formally joined the project in 2000. The first Top500 list was published in June 1993 and has been published twice a year ever since.

It rapidly gained acceptance and is now regarded as the most important measure of the high performance computer market around the world.

Meuer says: ‘By 1993 it was difficult to define what a supercomputer was; there was already a Cray 1 machine in a museum in Germany. What we needed was an automatic measure from a benchmark, so when a faster one came in the slower one dropped out. Using Linpac was criticised, but after 15 years I would claim it was a wise decision because of the success of the Top500. Linpac allowed us to make an accurate measurement, rather than relying on manufacturers. People say they can cheat us, but they can only cheat us once. ‘I think the success of the Top500 is that we foster competition between countries, manufacturers and even computing sites. At the beginning of the Top500, Germany was always ahead of the UK, but now the UK is ahead. There was always a story of “the Japanese danger”, because US sites thought that Japanese sites would take the lead, but that turned out not to be true. Japan is actually below the UK and Germany and well below the US. The two leading companies now are IBM and Hewlett Packard; they were not represented at all in the first list.’

Meuer can demonstrate how positions in the list can generate a commentary about the HPC market, with companies coming and going and surprising new entries such as Dell replacing the old stagers like Cray and Silicon Graphics. Intel has more than 70 per cent of the processors in the Top500 list – in fact Intel’s chief technology officer is a keynote speaker at the conference, something unimaginable when the Top500 was created.

Meuer retired from his position as director of the University Computing Centre in Mannheim in 1999, assuming the position of Emeritus Professor and founding his own consulting company – Prometeus. This is the company that promotes the Top500 and the research work is financed entirely from banner advertising on the Top500 web site. This has given him more time to concentrate on his first love, organising the International Supercomputing Conference.

He says: ‘Maybe I have more to do now than when I was at the university. I like this, I don’t just want to live in the garden and cut my tree. I want to be involved in this exciting high performance computing business, I love it and its not really work for me.

‘At the beginning of the conference, I ran it with my secretary and my wife helped me. Now I have 10 people working with me, including my two sons, otherwise it would be impossible. We have a big exhibition with 85 exhibitors last year and we hope to grow it. I am still responsible for the conference programme and this keeps me busy. But fortunately I have lots of friends in the field and so it’s never too difficult to get them to accept my invitation. During the conference I am still the general chair and everybody is coming to me when things go wrong. I attend all presentations from the beginning to the end, except of course when we have parallel sessions.

‘In the early years I knew absolutely everybody, but it’s becoming more difficult to know everyone at the conference. Of course I like to say hello to everyone but sometimes I do miss some people and I apologise.

‘I am surrounded by younger people, but I am not the oldest person there. Last year we had a poster presentation from a professor from Chicago who is 90 years old.’

Meuer is determined to keep up his involvement with the conference he started all those years ago and, as far as those attending are concerned, it would not be the same event without the warm greeting and access to the contacts book of Meuer at its centre.

Hans Meuer CV

Education 1962 Graduated in Mathematics, Physics and Politics from University of Giessen 1973 PhD in mathematics from the Rheinisch Westfälische Technical University in Aachen

Professional Experience 1962–1973 Specialist, project leader, group and department chief at the Research Center in Jülich, 1973–1999 Professor of mathematics and computer science at the University of Mannheim 1999 – Managing director of Prometeus

Feature

Robert Roe reports on developments in AI that are helping to shape the future of high performance computing technology at the International Supercomputing Conference

Feature

James Reinders is a parallel programming and HPC expert with more than 27 years’ experience working for Intel until his retirement in 2017. In this article Reinders gives his take on the use of roofline estimation as a tool for code optimisation in HPC

Feature

Sophia Ktori concludes her two-part series exploring the use of laboratory informatics software in regulated industries.

Feature

As storage technology adapts to changing HPC workloads, Robert Roe looks at the technologies that could help to enhance performance and accessibility of
storage in HPC

Feature

By using simulation software, road bike manufacturers can deliver higher performance products in less time and at a lower cost than previously achievable, as Keely Portway discovers