From Physics and Math in Moscow to File Systems at Stony Brook and IBM
July 13, 2021
Interview
Article shepherded by:
Rik Farrow
I don't recall when I first met Vasily Tarasov. When we talked, we realized that we had both attended HotOS 2011 in Napa, California. But I think we actually first met during a FAST workshop, likely while talking about one of the many articles that Vasily has co-authored and published in ;login:.
Rik: My first question relies on an assumption: that you moved to the US to attend university. Is that true?
Vasily: Yes, that is true!
Rik: Emigrating seems like a big deal to me. How hard was it for you? What was particularly hard about it?
Vasily: At that time — in the beginning of 2007 — emigration did not seem too hard to me. After receiving the university admission letter, the hardest thing for me was to decide — to go or not to go to the US. I was about to graduate and had worked for about two years on a interesting job that I really enjoyed (Linux kernel engineer at what's now Parallels). The team there was extremely experienced and I could learn much more by staying. So, the future looked bright without going to the US and spending years working on a degree. In the end, the deciding factor was: if I don't like PhD student life, I can always return; and if I don't go — I will regret to the end of my life that I had not tried.
Rik: How did you wind up choosing Stony Brook? Did you have a particular academic focus/area in mind? If so, why?
Vasily: Ha, I always smile when someone asks me this question. And that's because Stony Brook was a complete and total accident! Several of my friends decided to apply for grad school in a US university. I tagged along purely out of friendship spirit. I was applying to Computer Science programs because that's the field I liked. I did not want to spend too much money on applying to too many universities so selected four universities from four different quartiles in some university ranking. Stony Brook happened to be my accidental (and lucky!) choice.
Rik: How did you wind up with your advisor? Did you have any choice in the matter?
Vasily: I actually started with a different advisor than Erez... But, things did not go well with my initial choice. So, I switched to Erez after my first or second semester, after taking his OS class. I am grateful to this day that he agreed to take me despite possible inconveniences!..
Rik: How important was getting papers published in proceedings? Was presenting papers difficult for you? What was your favorite paper (if you have one)? Or papers.
Vasily: When I joined grad school, I had no idea about the process and value of publishing. Quickly, however, it became clear — I have to publish in several high-quality conferences to graduate. Presenting papers was not too difficult for me (I like teaching and explaining things to any audience), but getting my first publication in and even more so the time before my first publication was quite difficult. We were submitting papers and getting lots of rejections. This was the time when one starts to doubt his or her abilities, path, choices. However, as soon as our first paper got accepted — HotOS! — it gave me a huge boost of confidence.
Here are two papers that I really liked and frequently remember (quite biased choices):
- Scalability in the XFS File System, 1996 [1]
- Generating Realistic Impressions for File-System Benchmarking, 2009 [2]
I read the first paper when I joined the Stony Brook file system lab, and wanted to understand how local file systems work. That paper is very detailed about implementation, explains why they made many design decisions, and I couldn't find similar academic publications for ext2 or ReiserFS. When anyone asks about file systems, this paper is a classic example.
I like the second paper because it was presented at the first FAST conference I attended. I was specializing on file system performance and the paper has a very nice mix of math, theory and implementation. You could say I liked it mainly for personal reasons.
Rik: Why did you choose industry over academia?
Vasily: Well, I went to IBM Research because I could not decide between industry and academia. I think this is one of a few organizations where one can combine academic activities (research, publishing, mentoring younger researchers, community service) with very direct impact on products, services, and business, via understanding real customer requirements, defining products, designing and developing systems, supporting and servicing them. That's the balance I currently enjoy.
Rik: What areas are you working on now?
Vasily: In the last 2-3 years I, along with my wonderful colleagues, work on making file systems cloud native. We look into, for example, how to make existing file systems more consumable in hybrid cloud environments and what kind of storage do serverless platforms at the Edge require.
Rik: I notice that you've become a manager. How did that happen, as I thought that was unusual for someone working remotely.
Vasily: Unlike a lot of people, I was already working remotely when COVID appeared, shutting down offices. I had started working for IBM while my wife was a research professor at Stony Brook on Long Island. More recently, she found related work in the Los Angeles area, and we moved to be close. I don't mind flying into San Jose to visit the office, although I haven't been doing as much of that recently as I used to.
People knew I was interested in becoming a manager, and when a position in my group opened up, they offered me the job, even though I work remotely most of the time. I think that the pandemic worked in my favor here.
