Abstract
Conventional power management methods for processors may be reactive, adjusting frequency only after a computing device exceeds utilization thresholds. This reactive lag may result in increased latency for sudden workload bursts and allow low‑priority background tasks to trigger voltage increases. This publication describes a predictive frequency scaling method based on real‑time introspection of operating system workqueues. Implementing the method, a computing device may extract internal statistics (e.g., queue depth, task wait times) through event‑driven kernel tracepoints. The computing device can identify task priorities by resolving function pointers and incorporating application‑level telemetry. The computing device then generates a predictive workload index using time‑series forecasting models to anticipate future computational demand. Next, the computing device adjusts processor voltage and frequency levels based on this index before utilization metrics breach reactive thresholds. This predictive frequency scaling method allows for look‑ahead scaling for latency‑sensitive tasks and the enforcement of frequency ceilings for background operations, thereby managing performance and battery consumption.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
VV, Girish, "Predictive Workload Modeling and Voltage Scaling Using Event‑Driven Workload Queue Telemetry", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10300