The GÉANT Cloud Flow (GCF) platform allows to run large-scale workflows consisting of tasks. Tasks are typically embedded in Docker Container images, and GCF deploys these Containers to the most convenient GÉANT site offering a Container platform. The project goal is to extend GCF’s services with open-source software so that a user can not only run containerized-tasks, but platformed-tasks, i.e., tasks running within overlay platforms to leverage better the underlying cloud resources and make the workflow faster. For example, this extension would allow the execution of complex machine learning tasks within the Horovod distributed training platform, deployed by GCF as part of the workflow.
This pilot has been submitted by CNIT (National, Inter-University Consortium for Telecommunications, a non-profit consortium established in 1995, bringing together 38 public Italian universities to perform research. The PI for the project is Prof. Andrea Detti from the University of Rome Tor Vergata.
The Arrowhead Tools project is Europe’s largest project for solutions in automation and digitization for industry.
The GÉANT Cloud Flow (GCF) platform was deployed in the lightweight run-time digital twins installation segment of the project, based on machine learning technology trained with simulation and measurement data. The industrial partner in the project was ABB. This application is improving design of electrical motors.
How the workflow is executed in Cloud Flow
The European Platform for Sports Innovation
The Europen Platform for Sports Innovation (EPSI) is an academic group at a university in Finland collaborating also with CSC on AI. The workflow executes training for neural networks, i.e. machine learning and for analyzing video.
This pilot materialized out of the Arrowhead project and is similar in nature. This is an experiment using the ELMER FEM container at the industrial company Outokumpu, the largest producer of stainless steel in Europe.