Alan Nunes presented his thesis work, in the context of WP3 of this Associate Team, as a research poster at SC 2024, the largest conference in high-performance computing.
In this work, advised by Lucia Drummond (UFF) and Laercio Pilla (Inria), they have integrated performance and energy consumption capabilities to Flower (a Federated Learning framework) to capture information and enable better workload distribution decisions. They have proposed and implemented two optimal scheduling algorithms that optimize both performance and energy consumption, with a higher priority for one or the other metric.
These results have also been recently published and presented by Alan during SBAC-PAD 2024: https://inria.hal.science/hal-04690494v2
