Abstract
The present disclosure relates to a system and architecture for optimizing power efficiency in data centers running high fluctuation loads, such as machine learning operations. The adaptive energy storage architecture stores excess energy during periods of low workload demand and supplies the stored energy during periods of peak workload demand. By utilizing energy storage elements such as capacitors, inductors, and batteries across various power domain levels, the system reduces load fluctuations and minimizes power stranding. Inductors in series on the power path can act as a low-pass filter. Active load balancing through switches and batteries routes excess current to charge the storage elements.
Keywords: Data Center Power Management, Dynamic Power Routing, Adaptive Energy Storage, Machine Learning Infrastructure, Load Balancing, Power, Electricity.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
N/A and , "Dynamic Power Routing: Optimizing Ml System Efficiency Through Adaptive Energy Paths", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9963