direkt zum Inhalt springen

direkt zum Hauptnavigationsmenü

Sie sind hier

TU Berlin

Inhalt des Dokuments

An Interview with DIMA Alumnus, Prof. Chen Xu, School of Data Science and Engineering, East China Normal University (ECNU), Shanghai, China

Lupe

 

... both the Stratosphere system and the opportunity to work abroad peaked my interest. ...

 

Chen Xu holds a postdoctoral researcher postion at DIMA from 2014 until 2018 under the auspices of the Berlin Big Data Center (BBDC).

 

http://chenxuhomepage.vosi.biz/

 

 

How did you learn about DIMA and when did you join?

 

I first learned about DIMA at a VLDB Summer School held in Shanghai, China in 2013. At the summer school, Prof. Markl introduced the Stratosphere (today Flink) system and happened to mention that there were open positions in his group. Needless to say, both the Stratosphere system and the opportunity to work abroad peaked my interest. After his talk, I approached him and inquired about his research as well as the open positions in the DIMA Group at TU Berlin. Volker was patient, very kind, and he answered all of my questions. I joined DIMA in October 2014 as a Postdoctoral Researcher under the auspices of the Berlin Big Data Center (BBDC).

What did you work on while you were there?

While at DIMA, over the past 3.5 years, I have had the opportunity to conduct research on varying topics and mentored numerous Bachelor’s and Master’s Theses. I conducted research in “Efficient Fault-Tolerance for Iterative Processing on Distributed Dataflow Systems.” Initially, collaborating with Sergey Dudoladov (a former PhD student at DIMA/TUB), to implement an optimistic recovery strategy (for a particular algorithm) and illustrate failure recovery without the need to employ a checkpointing scheme. This effort resulted in a demonstration that was showcased at SIGMOD 2015 [1].

Subsequently, I collaborated with Markus Holzemer (a former DIMA Research Assistant) on “Fault-Tolerance for Graph Processing” and supervised his Master’s Thesis on this very topic. We proposed an efficient unblocking checkpointing approach as well as a confined recovery strategy at ICDE 2016 [2]. This effort was then followed by a study on machine learning workloads, which proposed a replica recovery strategy at TKDE 2017 [3], that extends our fault-tolerance approach (initially developed for graph processing) to include iterative data processing as well. Currently, I am mentoring Rudi Poepsel Lemaitre (a Bachelor’s student and former DIMA Teaching Assistant), who is actively working on demonstrating all of these recovery techniques using Apache Flink, which was founded in the DIMA Group. Most recently, we submitted a demonstration proposal to VLDB 2018 [4] based on our project.

In addition, I mentored Ye Lyu (a former Master’s student at DIMA) on “Distributed Approximate Quantile Computations.” Currently, I am mentoring both Pandu Wicaksono and Shibo Cheng (both IT4BI Master’s students at DIMA) on “Hybrid Failure Recovery” and “Adaptive Checkpointing,” respectively. Collectively, these efforts have the potential to positively impact our research and result in future publications.

How has your time at DIMA helped you to grow professionally? personally?

While at DIMA, I have grown both professionally and personally. I have authored several papers at top-tier conferences and journals, which greatly strengthen my publications list and raise my profile as an academic researcher. At the same time, I am steadily growing as a mentor to both PhD and Master’s students. At DIMA, I was an instructor in several courses, which enabled me to further develop my teaching skills. Furthermore, I collaborated with other research groups to prepare and submit a research proposal. I learned a great deal during this effort. Over the course of my time at TU Berlin, I met many top international researchers who visited the DIMA Group, in person, and I had the chance to exchange ideas with them face to face.

On a personal level, I was able to experience many new cultures that differ from my own and this enabled me to expand my horizons. During the course of my day to day life, I was exposed to diverse cultures: many Germans as well as an international group of people from around the globe. This exposure helped me to further improve my communication skills. Moreover, I not only enjoyed living in Berlin (an international city), but I was also able to travel across Europe.

What advice would you give to individuals who aim to conduct Postdoctoral Research at DIMA?

i)        Identify your research topic based on both your past experiences and current DIMA research initiatives. Your research will contribute to the success of the group and the group’s standing will help you to promote your research.

ii)       Publish high quality papers at top-tier venues. If you aim to pursue an academic position in the future, your publications are of key importance. Publications at low ranked venues are not as valuable in your pursuit to secure a position at a good university.

iii)     Participate in teaching activities, but try not to overwhelm yourself. Develop your teaching skills. This will help you when applying for a professorship.

iv)     Collaborate with PhD and Master’s students. Mentor PhD students and recruit academically strong Master’s students, who have a shared interest in your research. Hire them as RAs (research assistants) or supervise their theses. If you guide them properly, they will be able to promote your research.

v)      Learn to write research proposals. Postdoctoral researchers must develop their grant writing skills. It is very important for their future careers in academia.

Where will you be going to next?

I will be joining East China Normal University (ECNU) as an Associate Professor in the newly formed School of Data Science and Engineering in Shanghai, China.

References

[1] “Optimistic Recovery for Iterative Dataflows in Action,” Sergey Dudoladov, Chen Xu, Sebastian Schelter, Asterios Katsifodimos, Stephan Ewen, Kostas Tzoumas, and Volker Markl, ACM International Conference on the Management of Data (SIGMOD) 2015.

[2] “Efficient Fault-tolerance for Iterative Graph Processing on Distributed Dataflow Systems, Chen Xu, Markus Holzemer, Manohar Kaul, and Volker Markl, The International Conference on Data Engineering (ICDE) 2016.

[3] “On Fault Tolerance for Distributed Iterative Dataflow Processing,” Chen Xu, Markus Holzemer, Manohar Kaul, Juan Soto, and Volker Markl, IEEE Transactions on Knowledge and Data Engineering (TKDE) 2017. 

[4] “Fault-Tolerance for Distributed Iterative Dataflows in Action.Chen Xu, Rudi Poepsel Lemaitre, Juan Soto, Volker Markl. Submitted to Very Large Data Bases (VLDB) 2018.

Biographie

Chen Xu is currently an associate professor in the newly formed School of Data Science and Engineering at East China Normal University in Shanghai. Most recently he was doing postdoctoral research with Prof. Dr. Volker Markl at Database Systems and Information Management (DIMA) Group, Technische Universität Berlin, as a senior research associate supported by Berlin Big Data Center (BBDC) project. Chen got his PhD degree in 2014 from Institute for Data Sciences and Engineering, East China Normal University, supervised by Prof. Aoying Zhou and also worked with A/Prof. Minqi Zhou. From Nov 2011 to Jan 2012, Chen severed at ECNU-MediaV Joint Lab of Computational Advertising as a research intern. From April to July in 2011, Chen studied as a visiting student at Data & Knowledge Engineering (DKE) Group, The University of Queensland, hosted by Prof. Xiaofang Zhou and Dr. Mohamed A. Sharaf. In 2009, Chen received the Bachelor Degree from Department of Computer Science & Technology, Hefei University of Technology. His research interest is large-scale data management.

Zusatzinformationen / Extras

Quick Access:

Schnellnavigation zur Seite über Nummerneingabe

Auxiliary Functions