.png){kind=icon}
.png){kind=icon}
Return
Power-Aware Uncore Frequency Scaling Based on Throughput Saturation
.png){kind=icon}
OA
AI
K
H
M
DOI：10.1109/tnsm.2026.3716735.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Power-control mechanisms are essential for web-based processing applications to minimize energy consumption without causing excessive tail latency degradation. However, several challenges remain, including the coordinated control of core and uncore components and the cost of application modifications. Therefore, we propose a control framework for the uncore frequency alongside existing techniques for CPU core frequency scaling. This framework adjusts the uncore frequency on the basis of the server’s throughput state and does not require any application modifications. Evaluations of a web application show that, under the evaluated conditions, the proposed framework suppresses the extreme tail latency degradation and energy increases observed with techniques used for comparison. It also shows a more balanced trade-off between energy and latency.
Keywords:
CPU uncore
DVFS
energy-saving
machine learning
Journal
IF:
5.4
Papers:
509
Citations:
9.2K
Organization
Cited Papers