My journey to the west: Database research at TU Munich

The starting point of my current career

For the past few years, I have occasionally received messages about how to start research in database management systems (DBMSs). Those messages take me back to when I was working on user-facing application engineering, when every low-level system – such as OS kernels or networking – seemed so fascinating and complex. Back then, I felt that only the most intelligent people could understand and build such systems (trust me, a website is similarly complicated to an OS 😄), hence being able to do so on my own would somehow prove my intellectual worth. While that sounds cheesy (though not to me at the time – shameless self-humiliation here), it was actually the main reason I decided to focus my career on system research. And among all the possible systems, I chose DBMSs – for reasons that are just as cheesy.

Why research?

Academia is intimidating: I have to teach — once handling a class of over 100 students at TU Munich, which was both terrifying and fascinating; on top of that, I also have to present my work in front of leading scientists (e.g., Phil Bernstein). Given all that, why did I choose to do research in the first place? The key reason is simple: there is no DBMS job in Vietnam that would pay me ransomly (and please, don’t mention VNG 😅💢). My thinking back then was straightforward – if I couldn’t make much money doing what I loved, I might as well aim for glory instead (me 8 years ago was so childish). Academia seemed like the fastest – or maybe the only? I don’t know to be honest – path that fit my personality. So, despite all the challenges, I went for it.

How system research looks like?

Of course, deciding to pursue research was just the beginning. The real challenge started the moment I tried to live up to that decision. Research unfolds in a sequence of steps, and at each one you are forced to make decisions with very little guidance. The first – and arguably the hardest – is to find an interesting problem to work on. Ideally, it should be something unsolved; otherwise, what’s the point? This is why people like to joke that many research problems are “made up.” In theoretical fields, that cliché sometimes is not too far from the truth. In practical areas like DBMSs, it’s closer to a 50-50 split. Projects like DuckDB, for example, show that database research can be both scientifically interesting and genuinely useful to society.

Beyond that, the topic you choose has to be something you actually enjoy – otherwise, the whole journey becomes much harder than it needs to be. This is also one of the key reasons why many PhD students burn out or even drop out, together with the usual side effects like peer pressure and health issues. In my case, I was lucky: my advisor is smart, so finding a problem was never the hard part. My job was mostly to tackle whatever he threw at me.

From there, the first real task is to learn the topic and dig into the related work. Because implementing an idea in systems research is incredibly time-consuming – see my advisor’s blog here – you really don’t want to skip this step. Otherwise, you’ll end up reinventing the wheel (as I’ve done more times 😤💢) and waste a painful amount of time.

After that comes writing the paper. If all you want is to pad your CV with as many accepted papers as possible, this part isn’t too bad – yes, even papers at top conferences like SIGMOD/VLDB aren’t too hard to write. The writing process only becomes brutally hard when you aim for real impact. Then you need all sorts of subtle tricks: framing, narrative flow, manipulation, even a bit of marketing. And after doing all of that – plus teaching and some unpaid community service like reviewing papers – you finally earn a PhD.

In the DBMS community, which is small and tight-knit, having 3-4 SIGMOD/VLDB papers during PhD is often enough to get you a postdoc almost anywhere in the world. This is because everyone knows everyone else and how good they are.

Summary

And that’s the story. Looking back, my path into DBMS research wasn’t all smooth – it was a strange mix of naïve ambition, culture shock, and the intimidating thrill of suddenly standing and collaborating with some of the brightest minds in the field. Over time, I have learned that research is not a test of brilliance, but a long, messy process of trying, failing, and slowly figuring things out. And somehow, despite the chaos, I’m still here – still curious, still learning, and enjoying the ride far more than I ever expected. Honestly, I am living the best years of my PhD. If anything, I hope my story encourages more people to jump into systems research – especially DBMSs. It is a lot of fun, just… with slightly less money than one might hope.