When I entered grad school, I broadly knew I wanted to do work in systems and architecture. I preferred to build systems over something more theoretical like machine learning, and I liked thinking about how to make a system run faster, or consume less memory.
When I started working with Remzi and Andrea on systems research, I had no idea how to do research, so picking an area was not even something I thought about consciously. I took baby steps working with more senior grad students and helping out with their projects.
It was while helping out on one of these projects that I found I liked thinking about file-system consistency. At that point, I simply did not know that you are supposed to choose a topic based on its potential for growth, depth and so on. If I had known all that, it is likely I would have chosen to work on one of the current "hot" topics. But I didn't, so I chose to work on the classic problem of file-system consistency.
But it didn't work out too badly since I was able to get two good papers out of it. These papers form the core of my PhD thesis, and I had a lot of fun working on these projects. As Ben Y. Zhao points out, the key to happiness is finding your own research style and goals, and I loved disproving long-held assumptions about file-system consistency (e.g., "file systems cannot maintain consistency without ordering", "file systems cannot maintain consistency without flushing the disk cache").
After a couple of years of doing full-time research, I'm more aware of the strategic and tactical things to consider before picking a topic. I think I'm making more informed decisions on what to work on.
As to why I love my area and my topic: system research forms the bedrock of many of the exciting things going on right now. Google/Amazon/Facebook all use extremely interesting systems internally, and have introduced great innovations in operating systems and distributed systems.
With the explosion of data that is taking place, its becomes important to ensure the integrity of the data, while allowing systems to run at peak performance. This is a hard, interesting problem, and one that fundamentally affects the millions of people who use these systems.
Read other answers by Vijay Chidambaram on Quora:
- Should I take an unfunded MS offer in a top CS school like CMU or Stanford or a fully funded direct PhD offer at a slightly lower ranked program like at UCSD or UMD College Park?
- I am in interested in research, shall I consider MS in IITs?
- How should I choose my research specialization for a master's program in systems area?
from Quora http://ift.tt/2eDqlRi
No comments:
Post a Comment