System and method to enable prediction-based power management

A computing device comprising:

• a non-transitory storage device having embodied therein one or more routines configured to enable prediction-based power management of the computing device;
• and
• one or more processors coupled to the non-transitory storage device and configured to execute the one or more routines, wherein the one or more routines comprises:
• an idleness information recording module, which when executed by the one or more processors, through at least one power management agent, non-intrusively observes and records usage and idleness information of the computing device so as to enable updating of an idleness database storing information based on when the computing device is idle, wherein the at least one power management agent is operatively coupled with a power management server through a communication channel;
• a learning module, which when executed by the one or more processors, using a neural network operatively coupled with the at least one power management agent, conducts deep learning of one or more idleness patterns of the computing device, the one or more idleness patterns being analyzed based on machine training conducted by the neural network on the idleness database;
• a prediction module, which when executed by the one or more processors, predicts future idleness of the computing device based on the deep learning of the one or more idleness patterns;
• a prediction-based lower power state transfer module, which when executed by the one or more processors, directly transitions the computing device to a lower power state based on the predicted future idleness;
• and
• an interface that communicatively links the computing device to other computing devices each containing its own non-transitory storage device and one or more processors executing the one or more routines,
• wherein the at least one power management agent (i) determines an optimal non-intrusiveness power configuration of the computing device and for each of the other computing devices based on the observed usage and idleness information of each computing device, and (ii) identifies optimizations of resource usage of each of the computing devices, and
• wherein outlier idleness patterns are modified using power actions timeouts before providing data into the neural network for learning and prediction.

A method for prediction-based power management of a computing device, the method comprising:

• non-intrusively observing and recording, using at least one power management agent, usage and idleness information of the computing device to enable updating of an idleness database storing information based on when the computing device is idle, wherein the at least one power management agent is operatively coupled with a power management server through a communication channel;
• conducting, using a neural network operatively coupled with the at least one power management agent, deep learning of one or more idleness patterns of the computing device, the one or more idleness patterns being analyzed based on machine training conducted by the neural network on the idleness database;
• predicting a future idleness of the computing device based on the deep learning of the one or more idleness patterns;
• directly transitioning the computing device to a lower power state based on the predicted future idleness;
• and
• communicatively linking the computing device to other computing devices each executing its own method for prediction-based power management,
• wherein the at least one power management agent (i) determines an optimal non-intrusiveness power configuration of the computing device and for each of the other computing devices based on the observed usage and idleness information of each computing device, and (ii) identifies optimizations of resource usage of each of the computing devices, and
• wherein outlier idleness patterns are modified using power actions timeouts before providing data into the neural network for learning and prediction.