Justine Sherry

Justine Sherry

Women@SCS conducted an interview with Justine Sherry, Assistant Professor in the Robotics Institute.

Women@SCS
So first I was wondering if you could tell me more about your background and your journey before coming to Carnegie Mellon.

Professor Sherry
Yeah. So I went to the University of Washington for my undergrad, and I was an International Studies major. My roommate was a Computer Science major, so she talked me into taking Programming 1 just for fun and I wound up switching to CS. If you’re not in this special program, you have to do a thesis, and so I went around looking for a thesis project in computer science. At this time, there was this grad student I knew because he had a radio show on the student radio station. I really liked his music, but it turned out he was also working on this project to measure the internet. He was telling me, “I always thought that people surely know how the Internet works, like, humans built it, right? We should know how it works, and it turns out that the internet is made up of over 40,000 networks. They call them autonomous systems and the only reason there’s any such thing as internet is because all of these networks worldwide just decide to connect to each other.”

Women@SCS
Oh, that’s crazy.

Professor Sherry
So no one organization controls it, and that means that nobody actually really knows how big it is - nobody actually knows exactly what path a packet is gonna take to get across the network. It all just sort of self-assembles and no one controls it. And so he was working on these projects to map the internet, and I was really interested in that. So I wound up working with him to do my senior thesis. After that, I was hooked on research, so I went to Berkeley and got my PhD. Then, towards the end of my PhD, I had really enjoyed teaching and working with students (and of course I continued to really enjoy my research), so I applied for faculty jobs and I came to my dream school, Carnegie Mellon.

Women@SCS
Yes, that’s amazing.

Professor Sherry
Happy ending.

Women@SCS
So I’m actually kind of curious about this radio station person so -

Professor Sherry
Yeah.

Women@SCS
Was this person like also a CS major or was he -

Professor Sherry
Yeah, he was a graduate student.

Women@SCS
Oh, I see.

Professor Sherry
So for undergraduates who want to get involved in research the absolute best way to do it is to work with a graduate student, because faculty members are really busy.

Women@SCS
That’s so true.

Professor Sherry
But graduate students have a lot more time and they are a little better at remembering what it’s like to be an undergrad.

Women@SCS
Yeah.

Professor Sherry
Yeah, and so together with this guy Ethan, he really advised my whole undergraduate thesis. He helped me find a project of my own within his broader agenda of mapping the internet, and he’s actually still a really good friend of mine. He’s a professor at Columbia now.

Women@SCS
Oh, that’s crazy. That’s pretty cool.

Professor Sherry
Yeah. So I really liked the faculty I worked with as well, but the person who advised me day-to-day was a graduate student. So for undergraduates who want to work on research I think it’s often the best sweet spot to find a grad student to work with.

Women@SCS
Yeah, okay, that’s super cool actually! I also feel like in my experience doing research, a lot of times the professor that I was working with disappeared for several weeks. So I was mostly talking to the PhD and to the other grad students and they really, really helped me through my research project. I definitely feel like oftentimes it’s the grad students that really help you out. So my second question: What’s the biggest challenge you’ve faced in your career?

Professor Sherry
In the middle of my PhD, I had a lot of paper rejections - and that was really hard because I had these projects I had been working on for a really long time and I was really invested in them. I cared; I worked with them all the time and I was really excited about them. Then I would submit them for publication and they would get rejected. Not only would I be sad about the work, but I would also start to worry maybe I’m not good enough to get a PhD. Maybe I’m not smart enough to be here. Other people seem to publish papers, why aren’t my ideas good enough, why isn’t my work good enough? And so it was really hard for me to recognize that I could be smart enough to be a PhD student but that my project maybe still needed that next extra-special thing to get published. And it was really hard for me to learn, you know, just because my work needs more to get where it needs to be, it doesn’t mean that I’m not good enough. And I still struggle with that a little bit, like, nobody likes to have their stuff rejected. Eventually most of my research did get published, and so I managed to file a PhD and get a lovely job. But I think that was definitely the most difficult thing, both having the work rejected and also having the self-doubt that came from having the work rejected.

Women@SCS
I see. At that time, what helped you realize that you were self-doubting yourself?

Professor Sherry
Part of it was A, eventually the work did get accepted and so it wasn’t like I was dumber six months earlier. And so I came to realize that I had been good enough the whole time, it was just that the work needed to evolve more to get to the point where it was. But also I watched a lot of other people who I thought really smart, like, really amazing also struggle with papers getting rejected. I know one of my friends who’s maybe the most famous out of our friends in grad school, he has a paper that’s cited over 1000 times and is hugely influential, but I remember it got rejected the first 1 or 2 times he submitted it, too.

Women@SCS
Oh, that’s insane.

Professor Sherry
Yeah and he’s now a professor at another big, fancy school. With a little more experience I started to realize that rejections are just sort of part of the business. It doesn’t have a lot to do with who you are or how good you are. That said, it’s not really a good piece of advice to somebody who’s junior and just getting their first rejection. I can’t magically give them the experience of having survived a couple of paper rejections and come out unscathed.

Women@SCS
That’s true. So you mentioned that you thought that in the end you realized that your work just needed to be polished over. For you, what was that extra step?

Professor Sherry
My advisor brought in a collaborator to teach me a bunch of hacking skills. He’s like an amazing computer systems researcher in that his work has, like, really cool ideas and he has a tremendous amount of impact in the way that networking is handled in BSD systems. Working with him, I learned a bunch of skills that I didn’t have before and once I had those skills I started looking at the problem in a slightly new way and it helped give me insight into how I should be designing my algorithm differently. Bringing in new collaborators who know something that I don’t know, who can teach me something new, has often been the right step (at least for me) to become a better researcher.

Women@SCS
I see, so for our last questiom, how did you become interested in cloud computing?

Professor Sherry
When I was at Berkeley, this was in the big heyday of MapReduce, and it was a really big thing. People were starting to talk about warehouse-scale computing and operating systems for clouds, and there were all these things that they were taking for granted that we didn’t have in networking so networks at the time were about these fixed boxes. You would buy these boxes and they would do one thing. You would buy a WAN optimizer and it would compress traffic. Or you would buy a firewall, and it would filter traffic and you plug in the box. And if you want two firewalls, you buy two boxes and you plug in two boxes. Everybody down-the-hall working on cloud computing was using software. You’d buy racks of general-purpose computing and you run code on them and maybe you don’t even buy the racks. Maybe you rent them from Amazon and you run code on them. So I wanted all the nice things that cloud-computing people had, the ability to run software and change what the software is doing just by changing a few lines of code. I wanted that for networks because I was looking at firewalls and caches and proxies - all of this stuff in a box, where you couldn’t just change the code. You had to like buy all these boxes and wire them together. That just seemed crazy.

Women@SCS
I see, and so did you become interested because you saw it as a source of a problem or a source of improvement?

Professor Sherry
I saw it as a solution to problems we had in networking. In networking it was really expensive to deploy new equipment because you had to buy lots of new boxes. It’s much cheaper to buy generic compute equipment, to buy generic servers.

Women@SCS
Right.

Professor Sherry
It was really hard to manage because every one of these boxes had their own administrative configuration. They had their own policies. Even if you just wanted to know, like, “What’s the uptime, what’s the CPU utilization,” they’d have a different tool whereas in cloud computing- like you know it’s just Linux, right? You just like look at top, you can use top for any application. So all I wanted was to take all these problems up into this. It was complicated. It was hard to manage. It was really expensive and I wanted to take those problems we had in networking and solve them with all of the things we already knew how to do in cloud computing.

Women@SCS
Oh, I see.

Professor Sherry
The thing that made it a little bit tricky is that networking workloads are very different than cloud-computing workloads so packets need to be processed at, like, many millions per-second. We did have to redesign networking systems so that we could run them in software, run them at a cloud provider that manages your infrastructure for you, and we wanted to be able to do all of that without suffering performance loss. We still needed to be able to forward at millions-of-packets-per-second and it turns out that millions-of-packets-per-second is way, way more than you could normally handle, but we designed some solutions.

Women@SCS
Oh wow, so was that your thesis?

Professor Sherry
Yeah, my thesis was about taking metal-box infrastructure and moving them into the cloud and probably the trickiest part about it was figuring out how to make fault-tolerant.

Women@SCS
What do you mean by fault-tolerant?

Professor Sherry
If something breaks, then a backup’s gonna take over.

Women@SCS
Oh, I see. So how did you go from working on the Internet to cloud computing?

Professor Sherry
So after I’d done all this work in the Internet-measurement space, I went to grad school, but I didn’t quite know what I wanted to work on in grad school. I knew I still wanted to work in networking but what I didn’t know was exactly what area of networking I wanted to work in. Data-center networks were really popular at the time but I didn’t really wanna work on that. Congestion control is always popular, and I didn’t really wanna work on that. Software-defined networking was a thing, but for some reason I wasn’t too excited about that, either. Then I met Sylvia Ratnasamy at Berkeley, who eventually became my became my PhD advisor. She started talking about these devices called metal boxes, and she was really interested in working on them. I had read this article in Ars Technica, maybe like a month before, about metal boxes and how challenging they were and all the problems, and nobody had really talked about metal boxes. I hadn’t learned about them in my networking class. I didn’t know a lot about them, and they seemed really weird and really interesting. So when I met Sylvia and she proposed that we work generally on these metal boxes, they were something that just sounded cool and maybe part of why they were cool is that I just didn’t know much about them. So I went and decided to work with Sylvia, which is probably the best move I ever made in my career. She was an amazing advisor, and we went and worked on metal boxes.

Women@SCS
Oh that’s super cool. In terms of finding your advisor so how did you find Sylvia?

Professor Sherry
So basically when I applied to grad school I went to my undergrad advisors, the grad student I worked with, and two faculty, and they gave me a list of schools. I had no idea what I was doing, so I just applied to all the schools. Then when you get in, you just get to go meet people. It turns out to be, like, one of the funnest things in your entire academic career because you go to these schools, they buy you nice food, you get to meet all the other grad students, learn about all of the cool research, and get one-on-one meetings with every faculty member in your area.

Women@SCS
Boy, that’s impressive.

Professor Sherry
It’s really cool. We do it here. So every spring, we have all the students that we admitted come and visit us, and I have two days of back-to-back meetings with students who can ask me about my work, students here, life in Pittsburgh, and they get to do that with every major faculty member at all the schools that they get into.

Women@SCS
So in that case, if you’re meeting students back-to-back almost every day, is that feasible with that many applicants?

Professor Sherry
Yeah it’s only the students who are accepted.

Women@SCS
Oh.

Professor Sherry
And then they do a matching. So, I actually came here and met with Srini and then with Dave, who at the time, was working on this project called FAWN. I thought, “Man, FAWN, like, that doesn’t seem very exciting.” And Dave had great success with FAWN. I went to a bunch of other schools. I went to Berkeley and I met a bunch of different faculty members there, but for some reason, Sylvia was the one who just clicked. I really thought that the work she had done in the past was really cool. I thought that the kinds of things she wanted to look at in the future were interesting and also I really liked the way that she worked on problems.

Women@SCS
Right.

Professor Sherry
She sounded like a fun person to work with and so something clicked. I decided to work with her and, yeah, it was great.

Women@SCS
So would you say that your relationship with your faculty advisor was more professional or more friendly?

Professor Sherry
It was friendly-professional. It was. I mean we worked together so it was a professional relationship but, like, I know her daughter and we would oftentimes go out, as a group, our research group. We would go out and get beers after a paper submission. Actually for my research group here, I have a happy hour every month because I think it’s really important to be friendly with the people that you work with.

Women@SCS
Wow, you sound like the best PhD advisor ever.

Professor Sherry
I only have one student and she was just here. You met her. So maybe she can tell you how mean I am for making so many changes to her slides. She just gave me a really good talk. I learned so many things from her practice talk. There were all these things that I didn’t know that I thought I understood and now she was explaining it, I was, like, “Wait, I didn’t know that, like, why is it doing that?” So she just taught me a ton of stuff, it was great.

Women@SCS
Oh, wow. Oh, I also have one last question. After you got your PhD and you were looking for a faculty job, how did you know that CMU was the right place for you?

Professor Sherry
Oh, gosh.

Women@SCS
Sorry. That’s a pretty loaded question.

Professor Sherry
Yeah, well, one thing was that my husband’s also a computer scientist. So he’s also on the faculty and that was one big and important thing in our choice that I was only going to go to a school where we would both have jobs. I had a couple of offers, and we narrowed it down to the schools where we both had offers and then, let’s see, CMU is really friendly department and people are super-nice here and, like, know each other. There are many departments where faculty are sort of closed off. Every lab is its own group, and CMU isn’t like that. People know each other from all sorts of different groups. Pittsburgh is also actually a really, really great city. I wanted to live in a city but I didn’t wanna live in a big city, like, I didn’t wanna live in New York.

Women@SCS
Oh, no.

Professor Sherry
And so Pittsburgh seemed to be a sweet spot in terms of places to live and CMU is just a top-ranked school. We would have really good students here.

Women@SCS
That’s true. Compared to, like, would you say that other universities that you got offers from, do you feel like they were less friendly than Carnegie Mellon’s department? Was that the main issue?

Professor Sherry
Not that they were less friendly, just that some faculty, like to run their research group and, like, they have a lab and everybody comes in the lab and maybe within that research group they’re really friendly but you might not know who works down the hall from you.

Women@SCS
Oh, I see.

Professor Sherry
And CMU somehow goes out of its way to make sure that, like, I see Jan Hoffmann all the time and he’s in PL, right?

Women@SCS
Oh.

Professor Sherry
Jean Yang and I like hang out, like, on the weekends. That’s really nice. And a bunch of the students who I interact with are also theory students. Like I have a theory student who’s collaborating on a project with me.

Women@SCS
That’s so cool actually.

Professor Sherry
And so I wanted to be in a department that was friendly as a department rather than friendly as labs.

Women@SCS
That’s true. I think that wraps up our interview so far. Do you have any like last-minute questions, concerns or anything else you wanna say?

Professor Sherry
I do wanna review the transcripts to make sure I didn’t say anything dumb.

Women@SCS
No worries! Of course we will send you the transcripts before we put them on the website. Thank you so much for taking the time to interview with me!

Professor Sherry
Awesome, thank you so much for interviewing me!