https://rsd-ecole.cnrs.fr/prog/
Agenda
Lecture EnOSlib - Bruno Donassolo, Matthieu Simonin (slides)
- Tutorial EnOSlib - Bruno Donassolo, Matthieu Simonin
A Walkthough EnOSlib’s features on Grid’5000:
Manipulating the basic objects (Host, Network, Roles)
Acting on remote resources (run commands, pipeline of actions)
Discovering some of the observability tools shipped with EnOSlib
Working with several networks
Mixing the compute resource types (virtual machines, containers)
All the resources are located in the enoslib/ subdirectory of the hackathon repository.
Round Table: Platform and load models
Keynote - Philippe Bonnet (slides)
Round Table: Observation and Tracing
- Hackaton #1: Running experiments
- Lecture: Data analysis - Arnaud Legrand
Avoid uggly graphics (page 13)
Introduction to R and the tidyverse (dplyr, ggplot2) to summarize and visualize a series of measurements.
Central Limit Theorem, confidence interval, and important hypothesis (pages 1-49)
Dependent variables, Linear regression, and important hypothesis (pages 1-41)
- Tutorial: Data analysis - Arnaud Legrand
- Synthetic data sets
Curation and graphical verification
Summarizing data
Linear regression (modeling and parameter evaluation)
Best PhD prize GDR-RSD: Ahmed Boubrima
- Lecture: Design of Experiments - Arnaud Legrand (slides)
Randomizing inputs to avoid bias and sequences to avoid temporal bias (pages 5-19 and pages 50-63)
Parameter identification and experimental workflow
Deciding input parameters: what for ? (screening, model design, pameter selection, optimization, etc.)
Hackaton #2: Now, it’s your turn!
- Evaluating stability/reproducibility for:
Possibly update the experiment design.
Tutorial: Design of experiments - Arnaud Legrand
“Experimental” functions would be provided through a shiny app
Generating simple designs.
Identifying parameters
Round Table: Simulation/Emulation/Experimentation
Hackaton #3: Start comparing between groups
Lecture: Archive, identication, description and citation of source code for research software - Morane Gruenpeter (slides)
Hackaton #4: That’s all folks
Discussion on the following topics:
Reproducibility, at what cost ?
Representativity ?
…