- https://doi.org/10.1109/ithings/greencom/cpscom/smartdata.2019.00080
An Improved Dynamic Frequency Scaling Approach for Energy Saving Based RBF Neural Network
- Jul 1, 2019
- Deguang Li +5 more
Since DVFS does not consider predicting system behavior in the future stage, in order to improve efficiency of DVFS in fine-grained, we proposes a CPU utilization prediction model based on RBF neural network. Our model first collects five typical system characteristics related to CPU utilization during system running, then we use radial basis neural network to fit the nonlinear relationship between these characteristics and CPU utilization of the next period. According to the predicted CPU utilization, frequency scaling is performed to change frequency in real time. Finally, we evaluate our model with classical DVFS model by means of different characteristics of task set. Experimental results show that our model improves the prediction accuracy of CPU utilization and changes frequency scaling effect of traditional DVFS.
- # CPU Utilization
- # Characteristics Of Task Set
- # Frequency In Real Time
- # Frequency Scaling
- # RBF Neural Network
- # Dynamic Scaling Approach
- # Real Time
- # Neural Network
- # Frequency In Time
- # Dynamic Scaling