Abstract
The technology relates to a framework for designing the thermal layout of compute resource racks to proactively mitigate heat concentration zones and optimize cooling efficiency. The framework utilizes an artificial intelligence (AI) model, specifically a generative adversarial network (GAN), to automate the layout design process. The generative network is configured to ingest physical and operational constraints and produce a simulated layout. The initial layout estimation is then verified by a discriminator network, which acts as a neural network approximation of fluid dynamics, providing a closed-loop feedback mechanism to refine the generated result. The framework allows for the exploration of a large design space to find high-quality, optimal configurations of compute resource racks. The framework automatically provides recommendations for rack placement, eliminating the need for extensive physical understanding or on-the-spot solution vetting.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
N/A, "RACK THERMAL TOPOGRAPHY GANS", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9086