2003 to 2008

F. Amblard et al., MDA Compliant Design of SimExplorer A Software Tool to Handle Simulation Experimental Frameworks

"ergonomic interface for the specification of the experimental designs and the distribution of computations to launch the corresponding simulation experiments on parallel machines"

Principles

• Black box components
• Low coupling between the model and the design of experiments
• Chains of experiments

Limitations

• Distributed computing not included from the ground up
• No language to express design of experiments
• Very few applications

No wide adoption

2008

Computing power

• Personal computer // 1 to 8 cores
• Computing server // up to 50 cores
• Cluster // up to 200 cores
• Grid // > 2000 cores

The mined road to the Grid

How to make it simple ?

The DREAM project

Formalism

	A Task runs an executable,
	receives and produces Variables.
	Variables navigate from one Task to another by means of Transitions
	eventually doing Loops.

A bottom up process

Gilles: "Hey guys, I would like to lauch 2 000 000 simulations?"

The OpenMOLE team:

And OpenMOLE became scalable!

Jeremy "Hey guys, I would like to run my NetLogo simulations!"

OpenMOLE team: "What is a Net Lego?"

And OpenMOLE discovered the geographer's world!

The BioEmergences team "Hey guys, the biologists need to do data processing on the Grid and do not want to download OpenMOLE "

OpenMOLE team: "Biolo what?"

And OpenMOLE became accessible through the web

Where is OpenMOLE now?

Fully functionnal application

Many methods

Data processing	Optimization	Sensitivity analysis	Design of experiments	Calibration

Widely adopted

Geography	Economy	Food processing	Biology

Future

Advanced methods

• Integrate existing methods
• Design new methods

Focus on model

Collaborative platform

• Share data
• Share models
• Share explorations
• Share results
• Publish / reproduce

Address multi-modeling

• Compare several models
• Explore model familly