State-of-the-art tools for modelling
Paul Chapron
Guillaume Chérel
Mathieu Leclaire
Jonathan Passerat-Palmbach
Romain Reuillon
Scientists use naturally parallel methods daily on their laptop:
- Parameter estimation
- Sensitivity analysis
- Replication
- Monte Carlo simulation
- Data reconstruction
- ...
Execution on the same program with different parameters and/or datasets.
Purpose
- Provide state-of-the-art sampling methods, experimental design
- Tackle scalability using distributed computing
- Is non intrusive in the model logic
1 - Model?
Stuff that you can launch, taking inputs and producing outputs
Zero deployment approach
- User code is automatically deployed at runtime
- No prior knowledge of remote environment needed
- No installation required on any machine
Works with almost any language / plateform running on Linux
Direct sampling
Full factorial sampling (Grid sampling)
Random sampling (Monte Carlo)
Latin Hypercube, Sobol sequence
List of inputs (CSV file)
Replication
Parallel data processing
...
Model calibration
Model calibration progress
Profiles
Profiles output
Diversity search
Diversity search output
Collaboration
3 - Execution environment
Today
Local computer with multiple cpus
Distant computer through SSH
PBS (on ssh)
SLURM (on ssh)
Condor (on ssh)
SGE (on ssh)
OAR (on ssh)
EGI Grid (trough DIRAC)
Adhoc Desktop Grid
Grid Computing
6 juillet, 14h-18h
Inscription sur iscpif.fr