EuroHPC-prosjekt eFlows4HPC, Enabling dynamic and Intelligent workflows in the future EuroHPCecosystem

The eFlows4HPC project uses workflow technology to optimize computer modelling tools that need High Performance Computing (HPC) resources to more effectively solve new problems in industry and science. Because HPC systems can be complex and be set up in many different ways, from large supercomputers to cloud services, it is necessary to use technologies that allow scientific applications and industrial applications to work seamlessly on different systems and, at the same time, make them efficient. The workflows can help computer simulation tools with more efficient handling of big data, integration of simulation models with complex HPC systems, and management of Artificial Intelligence. In eFlows4PHC, scientists meet with developers that are specialists in the workflow development in order to optimize their computer models. eFlows4HPC works along three main Pillars: digital twins in manufacturing, climate modelling, and urgent computing in natural hazards. NGI is involved in the urgent computing Pillar on using workflows for tsunami forecasting. The tsunami forecasting, developed at INGV (Italy), is presently able to predict the impact of earthquake-generated tsunamis and related uncertainties faster than real time. Because the prediction includes the uncertainty, this forecasting scheme uniquely allows us to predict not only the most likely outcome of the tsunami once the alert is triggered, but also the probabilities of missed and false alarms. In this way, the forecasting tool has the flexibility to adapt to different design purposes and political choices. However, a shortcoming at present is making optimal use of the HPC resources for the entire forecasting chain from triggering to data presentation. Right now, only the tsunami simulations themselves are run at the HPC system. In eFlows4HPC, we will employ the workflows to move the entire forecasting chain from input to post-processing to the HPC system, and at the same time explore how the system can utilize AI to make some operations more efficient. In this way, eFlows4HPC may facilitate better, more efficient, and more flexible deployment of tsunami forecasting operations.

-

eFlows4HPC aims at designing and implementing a European workflow platform that enables the design of complex applications that integrate HPC processes, data analytics and artificial intellegence, making use of the HPC resources in an easy, efficient and responsible way as well as enabling the accessibility and reusability of applications to reduce the time to solution. The overall approach integrates three Pillars, Pillar I - Digital twins, Pillar II Climate, and Pillar III: Urgent computing. Different components from these three widely different user communities will be integrated into a common software stack, eFLows4HPC, that allow to reuse complex workflows in federated HPC infrastructures. The platform and software stack will be validated by use cases organised in the three Pillars. The Pillars will further develop methodologies and complex workflows that will be used as pilots for the industry, climate, and natural hazards communities represented by various EU Centre of Excellence. The Norwegian contribution will be primarily in Pillar III, and be related to implementation of workflows that enable faster than real time tsunami simulations. This will improve the effectiveness of the tsunami simulations, and help enable better tools that form the background for better warning procedures in the future. Integration of existing tsunami models with modern workflows will provide this enhancement. More detailed breakdown of the tasks are listed in the project Gantt Diagram. This Gantt diagram lists only the tasks of the Norewegian partner.