Lucas Nussbaum's Homepage

Associate professor at Université de Lorraine. Researcher in the MADYNES team, at LORIA laboratory, a joint laboratory with INRIA Nancy - Grand Est and CNRS.

Je suis Maître de Conférences à l'IUT Nancy-Charlemagne (Université de Lorraine). Mes activités de recherche s'effectuent au sein du LORIA, dans l'équipe MADYNES.

Contact information

Email: lucas.nussbaum@loria.fr

Bureau B136
LORIA - équipe Resist
Campus Scientifique - BP 239
54506 Vandoeuvre-lès-Nancy Cedex
France (map, other map)

Phone: +33 3 54 95 86 19

Research / Recherche

My research activities focus on experimentation for the evaluation of distributed systems in the context of High Performance Computing and Clouds. I work on the design of experimental testbeds (Grid'5000), emulation (Distem emulator), and Open Science and reproducible research.

Projects :

Grid'5000: Technical director. Member of the bureau, comité d'architectes. chief scientist of the Nancy site.
Software:
- Distem: DISTributed systems EMulator
- Kadeploy: scalable, efficient and reliable deployment system (cluster provisioning solution) for HPC clusters
- XPFlow: experiment control system for reproducible experiments

Publications

All my publications and talks since 2009 are available on the HAL Open Archive.

(Additionally, some older publications are listed on this page.)

Teaching / Enseignement

Je suis Maître de conférences à l'Université de Lorraine, affecté à l'IUT Nancy-Charlemagne. Je réalise une partie importante de mon service en Licence Professionnelle Administration de systèmes, réseaux et applications à base de logiciels libres (ASRALL).

Liens utiles

Listes de formations "Logiciel libre": April et AFUL

Useful Resources (sometimes outdated)

Code

Katapult: tool for automating the deployment of nodes using Kadeploy. homepage.
NISTNet on recent Linux kernels (2.6.26+). More.

The original NISTNet doesn't work on recent Linux kernels. There are some instructions (and a patch) to make it work on more recent kernels available here, but even this patch fails to work with more recent Linux kernels (apparently it worked until Linux 2.6.24). Here is an updated patch that is known to work with Linux 2.6.26. It applies to nistnet.2.0.12c.tar.gz, and the normal installation instructions work fine, provided that you install the necessary build dependencies (for Debian, at least x-dev libxt-dev libxmu-headers libxaw7-dev libc6-dev-i386 xmkmf xaw3dg-dev).

Another problem is that recent distribution kernels enable HPET, which breaks NISTNet. You need to recompile your kernel to disable it. Here is a patch against Debian's 2.6.26 kernel configuration that is known to work (a simpler diff might work too).

Note: if you are a Grid'5000 user, a sid-x64-base-1.1-nistnet image is available on Orsay (user lnussbaum). Log in as root, cd nistnet-3.0a, run make install, modprobe nistnet, and use cnistnet.
iperf's reverse mode. More.

iperf is a popular tool to measure network throughput. However, it can only be used to measure throughput from the iperf client to the iperf server. This is a problem when one wants to measure the throughput in both directions, and one of the hosts is behind a NAT/firewall. This patch adds a reverse mode, where the data is transfered from the server to the client, allowing to measure the download throughput. This mode is enabled by specifying the -2 option. The patch has been submitted to iperf's authors, but hasn't been included yet.

iperf's reverse mode patch, applies against iperf 2.0.4.
Linux kernel patch to enable/disable SCTP checksum computation via sysctl. More.

With SCTP, computing CRC32 checksums is known to add a signifiant overhead. Unfortunately, most NICs don't support offloading that computation. To be able to achieve Gigabit-grade bandwidth with SCTP with current hardware, it is required to disable SCTP checksumming. This is only possible by manually modifying the SCTP implementation. This patch adds a module option (no_checksums) that allows to easily disable checksums computation.

Note: This breaks protocol compliance. You won't be able to communicate with hosts where checksums are enabled.

Included in Linux 2.6.30
Bazimbra: script to rewrite of the text/plain part of Zimbra emails to make it readable. download

Links

On tools:
- Version Control with Subversion : the free online version of an O'Reilly book about Subversion. In particular, "Basic Work Cycle" in chapter 2 covers the commands you will use daily if you don't do anything too complex.
- Vim : Graphical vi/vim Cheat Sheet and Tutorial, Seven habits of effective text editing
- Ruby stuff :
  - Ruby off the Rails. Nice introductory article on IBM DeveloperWorks comparing Ruby & Java.
  - Programming Ruby, The Pragmatic Programmer's Guide. Free online version of the first edition (for ruby 1.6). The second edition (for ruby 1.8) can be bought online or borrowed in my office.
  - Why's (Poignant) Guide to Ruby : An online Ruby book for beginning coders which reads like an illustrated novel.
  - ruby-doc.org, with core and standard library documentation
  - rubydoc.info
  - Good tutorial
- GIT stuff:
- Internet Measurement Tool Taxonomy
- Programming UNIX Sockets in C - FAQ
- Bibtex Entry Types, Field Types and Usage Hints
- INRIA's categorization of publications for activity reports
On doing a PhD, writing your PhD, and steps to do after a PhD (in France):
- Présentation "La thèse" de Sébastien Mosser
- The illustrated guide to a Ph.D.
- Arnaud Legout's Recommendations for Ph.D. students
- A. Guénoche : De la thèse considérée comme un genre littéraire
- Guilde des doctorants, en particulier Guide du doctorant
- Après avoir candidaté Maitre de Conférences, j'ai rassemblé quelques conseils sur cette page.
- Opération Postes
- The Art and Science of Writing. How to write easy-to-read text?
Lists of conferences and call for papers:
- http://grid.hust.edu.cn/call/
- ETHZ's confsearch
- Networking Conferences Statistics: acceptance rate for many conferences
- Wikipedia's List of Computer Science Conferences. Also useful: Australian ranking for conferences and journals, ArnetMiner.org's ranking
- WikiCFP
- Journal of Universal Rejection (very prestigious journal, according to acceptance rate)
- Journal of Irreproducible Results (website down, Wikipedia page up)
- Improbable research: Research that makes people LAUGH and then THINK. includes Luxuriant Hair Clubs for Scientists and articles such as How to Write a Scientific Paper
Workgroups and mailing lists:
- ORAP - Promouvoir le calcul haute performance (mailing list)
- HotDocs -- lieu de réflexion et d'échange sur la formation doctorale
- GDR ASR (Architecture, Systèmes et Réseaux) du CNRS : asr, asr-forum, asr.rescom
- Groupe Calcul (mailing lists calcul et asr-calcul)
- Développement Logiciel à l'INRIA (et liste devel)
Teaching:
Indemnités Journalières par pays, ou sur cette page
English for software localisation. A list of common mistakes non-native speakers make when writing packages descriptions in Debian. Applies very well to other contexts too.
... and Piled Higher and Deeper and xkcd of course!