19:40 - 20:00
The need for computational resources keeps on growing from year to year, and while large scale HPC infrastructure keep on rising and strive to reach an Exaflopic performance there is also the need for smaller or specific, and more agile computing resources.
“HPC” is now everywhere, in all domains, from traditional high-end engineering to data analysis and AI. HPC is the backbone to all these computations but tended to be quite rigid and governed by long-term investments. Companies of all sizes now need access to HPC resources, in the most transparent manner and with a great flexibility… but not at any cost.
This need for flexibility must be reflected in the way HPC resources are deployed and accessed: from full on-premises to hybrid multi-cloud solutions, but also mixing HPC and AI workload on the same platforms. So not only do we need traditional HPC Centers capable of handling heavy parallel workloads, but also more on-demand infrastructures for peak usage, special requirements.
In this presentation, we will present a few projects where we tried to meet these requirements, with a review of the methodology and solutions we adopted.
CTO – UCit