Prof. Timothy Roscoe: Systems Research

Diese Seite existiert nur auf Englisch!

Timothy Roscoe received a PhD in operating systems design from the University of Cambridge Computer Laboratory in 1995. After a few years in the software industry in the US, he worked as Principal Researcher at Intel Research Berkeley Lab, Adjunct Professor of Computer Science at the University of California at Berkeley, and as a visiting researcher in the ERTOS program at National ICT Australia. An interview with Katja Abrahams.

May 2007

What is your main research area?

My main research area is systems: distributed systems, operating systems, and networking. Systems research works by building artifacts: one builds real systems in order to understand a problem space, and the kinds of tradeoffs that are possible. As a discipline, systems research pulls in theory from other fields of computer science, but it is always grounded in the real world and real systems.

What is the importance of research in distributed systems and operating systems?

Distributed systems are the basis for most of the computing services we use today. Web services, mashups, instant messaging systems, social networking and the Internet itself are all large distributed applications. The problem of how to deploy, run and manage them efficiently is one of the core issues within distributed systems. Operating systems are obviously the basis of most computer software, but looking at the structure of an operating system is very helpful to understand other systems, like web servers or game servers, because they include lots of principles developed inside operating systems.

What are currently the most challenging issues in systems research?

Right now is an interesting time for systems research. Computational power and network bandwidth are increasing very rapidly but speed of light is not and so latency - delays in data transmission - is becoming more significant. A lot of the systems building challenges are dealing with that imbalance.

Only a few years ago, a typical high-end work station had two or maybe four processors. In five years time you will typically see machines on the market with, say, 64 processor cores inside. Writing programs for these kinds of parallel computers is much more challenging than programming for conventional PCs. Moreover, the operating systems running on multi-core machines will look very different from Windows or Unix.

Networks are changing, too. People doing research in Internet architecture complain that since the Internet is so big, important and useful, it is actually very hard to innovate. I've worked on projects in the US like PlanetLab and GENI, which are aimed at creating an environment where new network models can be tried out for real, using virtualization techniques, and there's a lot of interest in systems like this in Europe as well. Systems research is definitely picking up momentum, people are feeling inspired to work in this area.

Which courses are you teaching this semester and what will you teach next semester?

This semester I am co-teaching the undergraduate course in networking with Roger Wattenhofer from D-ITET, and a graduate peer-to-peer systems seminar with Gustavo Alonso and Spyros Voulgaris. Next semester I will be teaching a network architecture  course together with Roger. He has a more theoretical approach than me, so I'm looking forward to an interesting collaboration. Apart from that, I may also teach a seminar in advanced topics in operating systems.

After finishing your PhD in Cambridge, you worked in the US software industry for some time. Why did you decide to come back to academia?

I used to work at the Intel Research Berkeley Lab, which is closely tied to the University of California at Berkeley which offered me a position as Adjunct Professor of Computer Science. It was a great time and I really enjoyed the interaction with the students. At some point I realized that I would like to try to move fully into academia, and Intel was very supportive of that. After a few months at the University of New South Wales in Sydney doing operating systems research, I came to ETH in January of this year.

What do you like about teaching?

For me teaching is a rewarding job. I like working with students because they do not take things for granted and ask questions which make me look at my own work from different angles. Being forced to explain things, I also get a deeper understanding about certain problems and things I thought I already knew. I particularly enjoy teaching together with somebody else who has a different background to me, since you end up combining good ideas. Here at ETH, for example, I can work with Donald Kossmann and Gustavo Alonso who both do database research - an area which usually does not have much interaction with operating systems or networking.

Why did you become a computer scientist?

Actually, I got my first degree from Cambridge in mathematics, but I had always been something of a hacker, writing code and building systems. When I finished my studies, the job alternatives seemed to be either an accountant or a sort of jobbing programmer. Instead, I decided to stay at University and move to computer science, which gave me much more opportunity to do interesting stuff. I've stayed a researcher in industry or academia ever since.

What do you think of ETH as a research center?

ETH has a very strong reputation and is a place with world class researchers offering unparalleled resources. It is very easy to fall into collaboration with other people here, especially other systems researchers, and to do big projects, which is not always the case at other universities. So far, I really enjoy being here! The students are very motivated and I am looking forward to building a really interesting systems group.

Is there anything special you would like to say to our students or future students?

The great thing about computer science is that you really get the chance to change things. Computers have a huge influence on the world and on how it works. When you look at all the ideas that have been commercialized by start-up companies in the last ten years, you realize that they have almost all have come out of universities. So, I would like to encourage ETH students to take advantage of the working environment here and to create great, innovative things.