•     

Adrien Luxey

Adrien Luxey

  • <name>.<surname>@irisa.fr
  • IRISA/INRIA Rennes–Bretagne Atlantique
    Equipe WIDE, Bureau E301
    Campus Universitaire de Beaulieu
    35042 Rennes Cedex, FRANCE

I work on distributed systems, with an emphasis on statistical solutions to enable robust services in unreliable environments.

As an intern

I first joined ASAP (now WIDE) in 2016 as an intern, supervised by Davide Frey and François Taïani. Based on Stéphane Delbruel’s work on content placement, I investigated the use of Youtube tags to predict the location of future content viewers.

You can read my not-so-well-written internship report on HAL.

As a PhD student

Since the ASAP team was so nice (even before we moved on to being WIDE) I came back as a PhD student since October 2016, this time supervised by David Bromberg. The thesis title is the following:

Towards new solutions to build large-scale distributed applications

First off, we take a lot of interest in the probabilistic dissemination of information among groups of poorly connected devices. Then, by aggregating local information to build statistical models, we can infer a distributed system’s future state. Such a self-aware distributed system can perform tasks more intelligently.

Scientific publications

I participated on three papers at the time of writing:

  • Yérom-David Bromberg, Adrien Luxey, François Taïani. CASCADE: Reliable Distributed Session Handoff for Continuous Interaction across Devices. 38th IEEE International Conference on Distributed Computing Systems (ICDCS 2018), Jul 2018, Vienna, Austria [hal];
  • Simon Bouget, Yérom-David Bromberg, Adrien Luxey, François Taïani. Pleiades: Distributed Structural Invariants at Scale. DSN 2018 – IEEE/IFIP International Conference on Dependable Systems and Networks, Jun 2018, Luxembourg, Luxembourg [hal];
  • Adrien Luxey, Yérom-David Bromberg, Fábio Costa, Vinícius Lima, Ricardo Da Rocha, François Taïani. Sprinkler: A probabilistic dissemination protocol to provide fluid user interaction in multi-device ecosystems. IEEE International Conference on Pervasive Computing and Communications (PerCom 2018), Mar 2018, Athens, Greece [hal].

The customs here at Wide is to list authors in alphabetic order (we relaxed that rule for Sprinkler). I did most of the development on Sprinkler and Cascade; Pleiades is primarily Simon Bouget’s work.

Source code

I did a lot of coding during my thesis: mostly gossip projects in Golang, packaged in Docker containers and controlled with Python. I am happy to share these projects, as I finally took some time to clean this mess my code used to be:

  • Gossip: Go project containing abstract structures shared by other gossip projects;
    • Cyclon: A Random Peer Sampling algorithm, still in Go;
    • StoryTeller: The sequence-sharing algorithm at the foundation of the Sprinkler and Cascade papers — still Go;
  • CSVLog: A Go API to log stuff to CSV that suits me;
  • Transport: A Go layer-4 network daemon;
  • Cyclon-test: An example pipeline of how I package my Go code into Docker images, and run many containers under the same network to test an algorithm. This is mostly Bash and Python as notebooks, because notebooks are so practical to craft plots.

Alas, this is poorly documented. Please ping me if you are interested in my code and need help!

Comments are closed.