Very large monolithic Exascale or near-Exascale HPC systems are emerging. But they are not operated in isolation: they are more and more entangled in a complex local environment of other subsystems such as storage or more specialised compute/data processing systems. They are also increasingly connected to external data sources or repositories. This session will explore the trends and perspectives along this path, with different possible angles: Is it still worth investing in very large monolithic systems? Do we want and need more diversity and flexibility and assemble more specialised partitions of resources? If so, what are the extended but local network requirements, and what software stack evolutions are needed? Will and can the computing centre itself become the “hybrid supercomputer”? How to improve connection also beyond the computing centre, ensuring better digital continuity, and some kind of federation, between centralised processing power and distributed sources or repositories of data?

Presented by