Joint Seminar: Towards Self-healing Climate Simulation Workflows

At any one moment in time, researchers affiliated with the Institut Pierre Simon Laplace (IPSL) climate modelling group, are running tens of global climate simulations. These simulations execute upon a heterogeneous set of High Performance Computing (HPC) environments spread throughout France. The IPSL's simulation execution runtime environment has recently been enhanced so as to support extended use cases. Such use cases include simulation monitoring, data publication, environment and simulation metrics collection, automated simulation control, self-healing behaviours. At the core of this enhancement is the Prodiguer messaging platform. The runtime environment now emits information, in the form of messages, for remote processing at IPSL servers in Paris. The remote message processing takes several forms:

• Persisting message content to database(s);

• Notifying changes in a simulation's execution status;

• Launching rollback jobs upon simulation failure;

• Dynamically updating controlled vocabularies;

• Notifying downstream applications such as web portal;

We will describe how the messaging platform has been implemented from a technical perspective and demonstrate the Prodiguer web portal receiving notifications.




13:30 h


Bundesstr. 53, room 022/023
Seminar Room 022/023, Ground Floor, Bundesstrasse 53, 20146 Hamburg, Hamburg


Sébastien Denvil, IPSL (CNRS), Paris, France


Reinhard Budich

Back to listing