Abstract
This disclosure describes an adaptive orchestration system that can dynamically route computational tasks to an appropriate backend, such as a local processor or a remote cloud service. The system can use a multi-objective optimization approach, balancing the semantic complexity of a given task, which may be estimated by a lightweight proxy model, against the real-time physical state of the device, including its thermal status, battery level, and network connectivity. This method of hardware-aware routing can proactively manage thermal loads and power consumption, which may improve operational efficiency and mitigate performance degradation associated with system-wide throttling on such devices.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Bhardwaj, Utkarsh and Awasthi, Shivank, "Adaptive, Hardware-Aware Routing of AI Tasks Based on Semantic Complexity", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10356