About - Research - Publications - Teaching
- Born
- 1990
- Degree
- Ph.D in Computer Science
- Organization
- Inria
- Position
- Research Engineer in Inria's Technological Development Service
- Contact Information
-
- - Email: luke.bertot at inria.fr
ICPS team, ICube laboratory, University of Strasbourg, 2014-2019
This work revolves around the execution of scientific/industrial workloads in
cloud environments.
The virtually unlimited resources offered by cloud operators could be leverage
for the execution of some scientific workloads. However combining the
provisioning of resources with the scheduling of tasks in variable environments
offers new challenges.
I have worked on integrating this variability in simulations to build tools for
studying and predicting the execution of batch jobs in cloud environments.
These can be used to :
-
Predict the cost and runtimes of workloads,
-
Study the impact of variability on existing systems,
-
Design new systems operating efficiently in environments where
variability is a constraint.
Resist Team and Grid'5000 technical team, INRIA Nancy--Grand Est, 2019-2021
As a member of the Grid'5000 technical team, I worked on tools and features to
help with large scale experiments and reproducibility:
-
I lead development on Grid'5000's (GENI-SFA) Aggregate Manager API, which
allows researchers using the Fed4FIRE European testbed federation to use
Grid'5000 resources.
(https://www.grid5000.fr/w/Fed4FIRE)
-
Developed Grid'5000 integrated Jupyterhub instance to facilitate deployment
of Jupyter notebooks on all parts of the Grid'5000 platform.
(https://www.grid5000.fr/w/Notebooks)
-
Designed and prototyped a tool to extract experiment metadata from the
Grid'5000 platform, to facilitate scientific data preservation.
(https://www.grid5000.fr/w/Grid5000_Metadata_Bundler)
RMS team, INFRES department, Telecom-Paris, 2021-2023
This work revolved on designing a distributed storage solutions for far-edge
clusters. These clusters, built from consumer hardware and small industrial
computers, are expected to be highly heterogeneous and operate in highly active
environments. As such, our storage solution needs to have as low an overhead as
possible, and must be reliable in cases of node failure or connectivity loss.
This work is a partnership with Cisco France.
Technological Development Service (SDT), INRIA Nancy--Grand Est, 2023-
The CLUSSTER+ project aims to bolster french industry and research by providing
a unified sovereign cloud infrastructure for AI and HPC. The project gathers
commercial entities (Eviden, OVHcloud, Qarnot, CS Group, ActiveEon) and public
institutions (Genci, IDRIS--CNRS, INRIA, CEA) to contribute expertise and
computing resources to the new infrastructure.
As a member Inria, I participate to the design of the new platform with the other
partners. I am also in charge of designing and implementing the software component
necessary for Inria resources to be made available to the new infrastructure.
+ : CLoud Unifié Souverain de Services, de TEchnologies et d’infrastructuRes
-
Leveraging Notebooks on Testbeds : the Grid’5000 Case
L. Bertot and L. Nussbaum
Computer and Networking Experimental Research using Testbeds (CNERT) workshop, July 2021
HAL
-
Implementing SFA Support on an Established HPC-Flavored Testbed : Lessons Learned
L. Bertot, L. Nussbaum, and D. Margery
Computer and Networking Experimental Research using Testbeds (CNERT) workshop, July 2020
HAL
-
Improving the simulation of IaaS clouds.
Ph.D thesis defended in June 2019
HAL
-
Improving cloud simulation using the Monte Carlo method.
L. Bertot, S. Genaud, and J. Gossa.
European conference on Parallel Processing (Euro-Par), August 2018.
DOI: 10.1007/978-3-319-96983-1_29
HAL
-
An Overview of Cloud Simulation Enhancement Using the Monte-Carlo Method. (poster paper)
L. Bertot, S. Genaud, and J Gossa.
International Symposium on Cluster, Cloud and Grid Computing (CCGrid), Mai 2018.
DOI: 10.1109/CCGRID.2018.00064
HAL
-
Méthode pour l'étude expérimentale par la simulation de clouds avec SCHIaaS. (in french)
L Bertot, S Genaud, and J Gossa.
Conférence d'informatique en Parallélisme, Architecture et Système (ComPAS), June 2017.
HAL
Strasbourg, 2014-2019
While working on my thesis I was a teaching assistant at a 2-year undergraduate
school in Strasbourg (France). IUTs have a selective application process to
limit classes to 100~120 students and have a strong emphasis on practical work
through tutorials (in groups of ~35 students) and lab sessions (in groups of
~15). As a TA most of my work as was in tutorials and lab sessions:
-
Intro to algorithms and programming (1 yeari of tutorials & lab sessions):
Students first contact with programming and an introduction to python.
-
Introduction to computing (1 year of tutorials & lab sessions): Basic
architecture of computers and an introduction to C.
-
Network architecture (1 year of lecturing; 5 years of tutorials and lab sessions):
Hardware layers of the OSI model and setting up a proper IP network.
-
Fundamentals of OSs (2 years of tutorials & lab sessions): Introduction to
the workings of OSs and the usage of Linux system calls in C.
-
Network programming (1 year of lecturing; 5 years of tutorials & lab sessions):
Software layers of the OSI model and building network reliant software.
Palaiseau, 2023
I intervened as a teaching assistant in one course of the École Polytechnique's
Bachelor of Science program. This course was held mostly online with students
advancing at their own rate and checking in regularly with the teaching team.
-
Introduction to Networks (1 year): Students first contact with C, network
protocols, TCP/UDP communications, and threading.