How did you first become interested in Computer Science?
At 17, I had never seen a computer and didn’t know anything about computer science. I was applying to universities when someone told me about this new computer engineering school in Greece, which was for the selected few who achieved the high grades needed to secure one of its few and highly coveted positions. I took it on as a challenge, applied and made it in without really understanding what I was getting into. At first I was regretful—after all, what I really wanted to study was chemical engineering, and wondered what possessed me—but the first few weeks as an undergraduate student at the Computer Engineering Department at the University of Patra in Greece were enough to completely turn this feeling around to absolute joy. I became totally fascinated with computer engineering and everything about it.
You received your BS and MS degrees in Computer Engineering in Greece, before earning your PhD at the University of Wisconsin—Madison. What does the computing field look like in Greece now, both in terms of academia and new tech startups?
The computing field in Greece, even more so in these troubled times for the country, is blossoming. When I was a student and later worked professionally in Greece, before pursuing my PhD, Greece was a land of opportunity: there was a great deal of development, and bright, open-minded people ready to thrust ahead with new technological ideas and grow the private sector. In the academic world, international Greek stars like Christos Papadimitriou, Mihalis Yannakakis, and Paris Kanellakis have inspired waves of young scientists in all fields of computing. Now the environment of work and education is difficult for everyone: strikes at universities and the disorganization of the public sector along with economic unpredictability undermine progress and make everyday life really difficult. Nevertheless, the innovative Greek spirit, paired with the Mediterranean passion for achievement, is a key recipe for success and a promising way to overcome the hardships.
I have huge respect for my colleagues in Greek academic institutions who, despite the constant practical problems, manage to conduct world-class research and publish excellent results at top conferences around the globe. Database research and theoretical computer science are particularly popular subjects in Greece, and we enjoy excellent results and candidates at all levels of education coming from Greek universities. Greek companies are omnipresent at top European Union project consortia and very well connected in Europe. I also see great startups, many of which start in Greece and move to Silicon Valley after their first round of financing.
Data collection is growing exponentially, and to manage and interpret the data, application developers need techniques which exploit the underlying hardware in order to “boost” system performance. How do you think the skill set of data management professionals will need to evolve in the coming years to keep up with trends in the field? How will storage in the cloud impact how we develop technologies to interpret data?
I was lucky to study and maintain a highly interdisciplinary nature in my work, spanning microarchitecture, storage devices, data management systems, and applications in several domains. Interdisciplinary research is more challenging, and in many ways one feels one is going against a very strong current: it requires a lot of time in meetings while building bridges amongst different teams, and open-mindedness and creativity in educational methods for young professionals. The field of data management today benefits particularly from an interdisciplinary educational and industrial approach as it sits naturally between the application (the user) and the computing platform. A data management system must deliver the right set of functionalities to the application while ensuring maximum exploitation of the underlying hardware and devices, not only for performance but, more importantly, for energy efficiency. Therefore, today’s data management professionals must be “vertically integrated”: they must understand fully the application requirements and connections to the data management software while at the same time optimizing the interaction and utilization of the hardware underneath.
Storage in the cloud greatly simplifies traditional data management because it decouples elasticity from performance and allows for practically efficient resource provisions and management. At the same time, the additional levels of abstraction limit the control the database system used to have at all storage levels of the data hierarchy. Nevertheless, the fact that the systems issues are taken care of at the cloud storage level enable the creation of many data management services by an ever-growing community of users. The net result is a huge variety of data management tools of all conceivable kinds developed on the cloud, with service guaranteed by a stack of service-level agreements (SLAs). The space is growing and many worry about the lack of convergence; I believe that the era of convergence in a commonly accepted “database system” is long passé and that multidimensional scientific and entrepreneurial activities define the new age of science and technology.
You have worked on computational databases for data-intensive medical applications, and new approaches in database management are being used in astronomy and other fields. What are some exciting examples of new research projects in the health or natural sciences that will be possible because of advances in data management?
I am fascinated by projects which have as a main goal to help people (like the work of Chris Ré against human trafficking). Although my principal line of work has been introspective computer science (bridging database systems to computer architecture), I’ve worked on non-computer science projects since 1999 when my (then) PhD advisor Yannis Ioannidis introduced me to a group of soil scientists. As a graduate student I had never had such a tight grasp of reality as when I heard firsthand that the lack of a fast system for prediction of runoff at cranberry bogs in Wisconsin could destroy the farmer’s crop. We built such a system which worked, and although I moved on to a different subject of research I was forever looking for more. I worked on Jim Gray’s project with the Sloan Digital Sky Survey, and also on simulating earthquakes with Dave O’Hallaron at Carnegie Mellon.
At EPFL I have been working with experimental neuroscientists who simulate cognition functions in the human brain. Through this channel, I helped build the Human Brain Project and started leading an infrastructural effort to federate all hospital data in a scalable manner in order to create a sea of medical information and, through integration, mining, and causal modeling, build biological disease signatures which uniquely identify different types of brain diseases. The project is enormous and incurs a lot of overhead; its ever-changing structure feels like quicksand at times. In addition, the problems are far from computer science research, and we have to hire engineers and do a lot of legwork with hospital IT teams to be let in and discuss the work. Nevertheless, the outcome will be tremendously impactful. The coupling of clinical measurements and biomarkers into unique disease identifiers is unprecedented and key for precision medicine, which has the potential to transform today’s statistics-based diagnosis into semi-automated personalized care.
Many researchers in the database community (Juliana Freire, Magda Balazinska, Val Tannen, Susan Davidson, Arie Shoshani, Peter Baumann, and many many others) do excellent work in scientific data management. Unfortunately, practical communication problems across interdisciplinary teams, as well as the actual nature of the contributions, make for fewer publications. This is a shame because there are countless innovations in various disciplines which are in large part due to the hard work of database people. In the future, I hope to work with the leaders in these fields to find ways to make this work much more visible in major database conferences.
What are you most looking forward to at the upcoming SIGMOD/PODS conference? Why is participating in the conference worthwhile for professionals in the field?
SIGMOD/PODS was the first “big” database conference I attended in 1997, when I presented to the audience the demo of my soil science system. I can still remember being lost and found at the same time: lost in a sea of people, all of whom I wanted to meet, but at the same time feeling right in my element. Since then I have been going to SIGMOD/PODS almost every year, and served and chaired its various committees, before being honored by the international community with my election as SIGMOD Vice Chair. I look forward to talking to my colleagues and to indulging in new and exciting discussions, to catching up with young students and my academic family, and of course to attending DaMoN, a workshop collocated with SIGMOD that I co-founded with Peter Boncz and Stefan Manegold 12 years ago. What I mostly look forward to, however, is to learn.
SIGMOD/PODS offers a huge spectrum of advantages for professionals in the field. The most important asset for any organization--educational or industrial--is its people, and SIGMOD is a great venue to find and recruit qualified people at all ages and at all levels. The conference attracts top academics and top industry, as well as representatives from all conceivable parts of the database sphere. Even people who are only marginally related to the field can attend the tutorials and panels and obtain a crash course on any data management subject that’s relevant today. ACM and the Executive Committee work hand in hand with the organizers all year to create an impeccable program and fine-tune all of its details. The energy at SIGMOD/PODS makes it a unique place to acquire concentrated knowledge, and emerge full of new ideas.