Interview with Aaron Schulman, 2014 ACM SIGCOMM Dissertation Award Winner
The Department welcomes comments, suggestions and corrections. Send email to editor [at] cs [dot] umd [dot] edu.
On Friday, March 8th 2014, Aaron Schulman (Ph.D. 2014) heard that he won the ACM SIGCOMM Dissertation Award.
Aaron sat down with us to talk about the Award, his Post-doctoral work at Stanford and what being a graduate student in the CS Department meant (and still means) to him.
Walk us through the day that you found out that your advisor nominated you for the award, and then let us know how you felt when ACM SIGCOMM let you know that you had indeed won the award.
Before I started working on the dissertation, my advisor, Neil Spring, and I wondered if the community would be as intrigued by the work as much as we were. When I found out that I won the award, I was pleased to find out that indeed, the community felt intrigued too.
What was the beginning of the research process like for you? How did you decide on your dissertation topic?
In the beginning, I focused on learning the mechanics of research. I needed to learn how to collect data carefully, visualize data creatively, and write text clearly. Luckily, I started graduate school with a patient advisor and a lab full of thoughtful graduate students who taught me these skills.
I decided my dissertation topic indirectly. First, I pursued the problems that excited me. Then, I searched for my internal compass that led me to work on these problems. I would not recommend this approach to new graduate students. Instead, I believe students should continuously try to classify the set of problems that interest them. Each time students come up with a new classification, they should make a plan for building a dissertation out of that set of problems.
When you started graduate school, did you know what you wanted to work on specifically? If not, how did you get to networks and network measurement?
I knew that I wanted to work with networks, but my advisor helped me find out that the specific passion I had was for network measurement. Looking back, I realize now that even as a kid, my friends and I were excited by observing the behavior of networks.
I had a friend in middle school whose father worked in IT. His father would bring home older networking toys from work for us to play with. At one point, he brought home an Ethernet hub, which featured a curious little light that said “Collision” over it. As I found out later, that light blinks every time two computers connected to the hub send a packet at the same time. However, at the time we did not know what caused the light to blink, so we ran several experiments to try and make it blink. We never figured it out, but those experiments were the beginning of my interest in observing interesting phenomena in networks.
Can you talk about your interest in hardware?
Software will always be my favorite part of computing. When you are solving a problem by writing software, you can explore many different solutions with almost instantaneous gratification. However, I have always wondered what magic went on in the circuits of the computer to give my software such magnificent power. For example, I knew I could write a few lines of code to play a song, but I wanted to know what magic went on in the circuits to make the speaker sing.
I believe I can trace this need to understand hardware from growing up with a father and a grandfather (on my mother’s side — yes, I got it from both sides) who spent most of their spare time playing with electronics. They would tinker with TVs, radios, VCRs, and even vacuum cleaners. Both of their basements were covered with bins of electronic parts and half-built projects. Neither had degrees in Electrical Engineering, but both had a passion for learning about it.
What was the last push to finish the dissertation like? What is your advice for people who are working on the last parts of their dissertation?
In order to finish the dissertation, I needed peel back the onion of my entire academic life. I had to learn my strengths then buildup my confidence in them, and identify my weaknesses then work around them. For instance, I found out that I speak better than I write. To work around this problem, I would walk around DC with a voice recorder and dictate each section of my dissertation. Along the way, I would stop at coffee shops and parks, listen to the recordings, write down the text, think of what I needed to say next, and repeat the process.
If a person is struggling with his or her dissertation, I would advise that they describe, to an objective third party, what he or she perceives to be relative strengths and weaknesses. The third party can be someone on your dissertation committee. Therapists are also a great resource, especially the ones with PhDs as they have written dissertations themselves.
You're now at Stanford doing a post-doc. With whom are you working and what are your current projects?
I am working with Sachin Katti; he and his students also work at the intersection between software and hardware. In particular, the project I am working on is developing programmable wireless networking for mobile devices.
Right now on your smartphone, most of the wireless networking is implemented in hardware. The advantage of pure hardware implementations is that they are small and more energy-efficient than software implementations. The problem is this: today's wireless network hardware is application specific, so whatever wireless networking applications your mobile device ships with, is what it will be thrown away with.
So what is the benefit of programmable wireless networking hardware in mobile devices? Instead of having a separate mobile device for each mobile communication system like we do now, a programmable mobile device could be programmed with a large number of wireless networking applications. The reason this is possible is that wireless networking hardware including walkie talkies, portable TVs, satellite radios, FM radios, and GPS receivers all perform the same general task, digital signal processing.
There are many problems that would need to be solved to achieve this vision, including developing flexible analog front ends and antennas. However, the first basic question I am looking to answer is: what is the tradeoff between wireless networking programmability and energy-efficiency?
Are you still doing work with anyone at UMD?
I am still working with everyone I was working with when I was a graduate student. In particular, Dave Levin and I are working together to apply some of the broadcast work in my dissertation. Also, Ramakrishna Padmanabhan, Youndo Lee, Neil Spring, and I are working on a massive study of weather-related last-mile link failures. Finally, my dissertation work led to a problem that Matt Lentz, James Litton, and Bobby Bhattacharjee are trying to solve.
What are the advantages to doing a PhD in CS at UMD? Why should a student choose to do his or her graduate work here? Why is it great to work in Networks?
University of Maryland is a unique school to study networks, because the professors follow a strict observational approach. First, they encourage all students to learn how to observe complex behaviors hidden in networks. Then, they guide the students in applying their observations to build new and interesting systems.
University of Maryland lies in the heart of the Silicon Valley on the east coast. Around Maryland, the combination of government agencies, government contractors, non-profit organizations, and now even many startups, provides for an extremely talented pool of PhD students. Also, having these organizations operating nearby makes it easier to partner with them on research projects.
Communication is a basic function of both computing systems and humans. As such, in the field of computer networks, an improvement to a network of computers can directly translate to an improvement in the lives of humans.