Supercomputers exist inside a growing scientific datasphere spanning from edge devices to HPC systems.

The Exa-AToW project aims at providing solutions for the efficient management of large-scale workflows composed of HPDA, AI, and HPC tasks that are distributed over a continuum of resources ranging from the Exascale, HPC, and Data infrastructures. (This is is illustrated in Figure 1).

Exa-AToW focuses on effective end-to-end solutions, at scale, by considering not only functional dimensions such as workflows and data logistics but also resource federation governance, cybersecurity, energy, and sustainability.

As a consequence, Exa-AToW faces these main challenges:

• How to include the Exascale system in the datasphere?
• How to ensure cyber-security at a large-scale on heterogeneous technologies?
• How to dynamically allocate distributed resources to accommodate for variations in the power cost?

To address its objectives Exa-AToW is developing a data-driven service-oriented architecture (collaborative system-of-systems). which integrates technologies with the corresponding governance and cybersecurity together with global optimization capabilities considering performance, energy and sustainability.

Exa-AToW main outputs are:

• Machine Actionable Data/Project Plan
• Application & Workflow Support Software Stack  (APIs, …)
• Governance and Cybersecurity Principles
• Examobile (Hard/Soft device for data logistics)

Exa-AToW gathers researchers from University of Rennes, Idris, Datamove/Inria, MEM/CEA, Phyteas.