Willy Zwaenepoel, currently Dean of Engineering at the University of Sydney, is renowned for his numerous contributions to operating systems and distributed systems. With degrees from Ghent University and Stanford University, he’s excelled in areas like distributed shared memory, Internet servers, and more. In addition to having reached fellowship status at both the Association for Computing Machinery (ACM) and the Australian Academy of Technological Sciences and Engineering (ATSE), he was recognized as an IEEE Fellow in 1998 as well. Zwaenepoel’s impact spans beyond academia and transcends into industry. For example, his TreadMarks distributed shared memory system was licensed by Intel and became the basis for the company’s cluster OpenMP Product. His many trailblazing contributions have earned him numerous awards, including the IEEE Kanai Award and the Eurosys Lifetime Achievement Award, solidifying his status as a leader and innovator in the field.
In honor of his many achievements, he has received the 2024 Harry H. Goode Memorial Award for, “…for seminal and sustained contributions to operating systems and distributed systems.”
Your extensive work spans various areas in experimental computer science. Could you share a specific project or contribution that you are most proud of?
I have indeed worked on many different topics, but the thing that I am most proud of is my contribution to community building. I am now at the twilight of my career, and when I look back, it is not my h-index or my however-many papers or this or that research project that warms my heart most.
Aside from the students and the postdocs whom I have supervised and the successful careers of many of them, the one accomplishment that I am most proud of is the Eurosys conference, of which I was a co-founder and the first program chair. At the time, in the early 2000s, computer systems research in Europe was very sparse. Only a very few institutions had achieved a degree of prominence in the field. The creation of a new conference was a huge gamble, and there were many skeptics, but there is now a thriving community of computer systems researchers in Europe. Most would agree that Eurosys greatly contributed to that.
Honor your colleagues’ achievements. Nominate Someone for a Major Award Today!
Your career reflects success not only in academia, but also in translating research into practice. Can you share what you’ve learned while bridging the gap between research and industry, especially in the context of your startups and collaborations?
The main lesson is that academia and startups are different worlds. I must admit, it took me a while to realize that. In research, you try to develop a new idea – and the novelty of that idea is key. You will not get your paper accepted unless you can argue that there is enough novelty in the work. In a startup, you try to find a solution for a problem. That may require new ideas or new technology, but the problem comes first. It needs to be a problem that keeps a good number of people with money awake at night, such that if you come up with a solution, they will pay for it.
I have advised several startups, mostly started by engineers, and you can already see from the first slide in their pitch deck whether they have it right or not. If their first slide is titled “Technology”, it is not likely to go anywhere. If their first slide is titled “Problem” and the second slide is “Market Size”, a fancy way of saying who is willing to pay what money for the solution of the problem, then there is hope. Candor requires me to admit that I have made that mistake myself, in fact, more than once.
How has your experience in startups influenced your approach to research and innovation, and how do you see startups shaping the future of computer science and technology?
It has to some extent shaped my approach. I like to work on things that I believe will have an impact in the real world, and I am not that concerned about whether I can publish the work at a top conference. I have published papers at top venues with substantial citation counts, but let me share this little-known fact. In the early 2000s, I published a paper in a good venue that had a good number of citations, but the more remarkable fact was that the paper was downloaded 50,000 times from my group’s website. A CTO of a Fortune 500 company told me he had made the paper required reading for all his reports. This is for me “dream come true” work.
That said, there should be room, especially in universities, for “blue sky” research which may not have an immediate impact or may never have an impact in industrial practice. I find that funding agencies around the world have become far too restrictive in this respect. For one thing, one never knows what is going to have an impact. Some very “blue sky” projects had, after all, a major impact. For another, if the work is scientifically sound, it trains students who will go on to successful careers, often in different areas.
In your opinion, how has the landscape of internet servers evolved since you started in the field, and what emerging trends or challenges do you foresee in this technology in the future?
The landscape is unrecognizable compared to what it was when I started working on Internet servers in the mid-90s. At that time, we were basically trying to retrofit hardware and software that were not designed to be Internet servers to do the trick. Now, all of that is in place. In many ways, the most impactful development was virtualization. Virtual machines were well known, of course, starting with the IBM-370, but their application was limited. The fact that you could run virtual machines on commodity hardware was a breakthrough. Virtualization is what made the cloud possible. I am happy that I was able to contribute some ideas to I/O virtualization.
You have been elected as a Fellow in IEEE, which is a major accomplishment, congratulations! How has being a part of this community influenced your career and research? Were you able to make strong connections? Do you have any advice for researchers just getting started who are interested in following a similar path?
One of the things that struck me when I moved from the US to Switzerland in 2003 was that fellowship in the learned societies in computer science was much less understood and valued in Europe than it was in the US. So, in my position as head of school, I set out to remedy this situation. Virtually every year I nominated a couple of people in my school. Most of them were successful, which shows that the inherent human capital was there. It just needed to be brought out.
I would put in a plea for young researchers to put effort into the community. I know there is a lot of pressure on them. A bit of pressure is good, but quite a bit of that pressure is unhealthy. The demands in terms of h-index, number of papers, and number of grants are pushing them to think near-term and to salami-slice their work into as many papers as possible. I am not sure what to do about it, but it is unhealthy.
Given your role as the dean of engineering at the University of Sydney, what initiatives or changes do you believe are essential in preparing the next generation of engineers and researchers for the evolving landscape of technology and computer science?
I hold a rather controversial opinion on this matter. There has been an enormous shift towards “job-ready” skills. Industry understandably wants fresh university graduates to be ready to do the job they are being hired for, but it is the task of the university to prepare a student for a lifetime of jobs, including but not only their first job. I agree that in the past university teaching was too theoretical, too far removed from industrial practice. The main goal of the university should, however, be to teach students to think. Some theoretical subjects and a good bit of mathematics are ideal for that purpose.
I went through a system where engineering was a five-year degree, the first two of which were essentially mathematics and physics. Only afterward, we picked a particular engineering field, in my case electrical engineering. In Ghent, where I did my undergraduate, there was no degree in computer science at the time. Since I switched to computer science in graduate school, I rarely used my electrical engineering training, but I am immensely grateful for the way especially those first two years trained me to think.
We see that about half of our students after five years work in a different field from what they studied. Some have concluded from that observation that what we teach is not that important. I do not think that is a correct conclusion. The right conclusion is the opposite: it is because we teach fundamentals that the students have a broad base that allows them to switch fields.
I am likewise a bit frightened by the move away from “core material” to things like “soft skills” and “leadership”. In my experience, leadership in a technical profession can only be acquired when one thoroughly understands the core technical underpinnings of the discipline. I am not saying that soft skills and leadership are of no use, of course not, but they are meaningless without a deep technical grounding, and that should come first.
Among the many areas you’ve contributed to, such as microkernels, fault tolerance, and graph computation, is there a particular field or technology that you find especially promising for future research and development?
I always try to look at areas where there is a fundamental technology shift. Right now, the most fundamental technology shift is the advance in large language models. In all honesty, I am a bit out of my league on the topic. I am also afraid that it is also going to be difficult for university researchers to play a leading role in this space because of the need for access to huge amounts of computing power to process massive amounts of data. I hope that I am wrong, but that is what it looks to me. Countries must invest in the computing infrastructure needed to sustain open research in this domain.
For my own work, I am intrigued by the advances in memory and storage systems. We have gone from the conventional “caches in front of DRAM” architecture to one in which there are a variety of different memory systems, some of them disaggregated. In the storage arena, new technologies allow access to nonvolatile storage at a fraction of the latency and at a multiple of the bandwidth of what was possible with earlier technologies. There is ample room for good computer systems research in this space.
More About Willy Zwaenepoel
Willy Zwaenepoel received the BS/MS from Ghent University in 1979, and the MS and Ph.D. from Stanford University, in 1980 and 1984, respectively. He is currently dean of engineering at the University of Sydney. Previously, he was on the faculty at Rice University and head of the School of computer and Communication Sciences at EPFL.
He has made fundamental contributions to experimental computer science, including groundbreaking work on microkernels, group communication, fault tolerance, distributed shared memory, Internet servers, I/O virtualization, graph computation, and storage systems. In each area, he is credited with breakthrough results.
He was elected IEEE Fellow in 1998, ACM Fellow in 2000, and ATSE Fellow in 2021. He received numerous awards for his teaching and research, including the 209 IEEE Kanai Award, the 2019 Eurosys Lifetime Achievement Award, the 2023 IEEE Outstanding Technical Achievement in Distributed Processing Award, and nine best paper awards at conferences including DSN, Eurosys, OSDI, Sigcomm and OSDI. He also received the 2001 Rice University Graduate Student Association Teaching Award.
He has been successful in translating his research into industrial practice. His TreadMarks distributed shared memory system was licensed by Intel and became the basis for Intel’s cluster OpenMP Product. He has been involved with several startups including iMimic (acquired by Ironport/Cisco), Bugbuster (acquired by AppDynamics/Cisco), Nutanix (Nasdaq: NTNX), and Grainite (acquired by MongoDB).
