×

Methods of achieving cognizant power management

  • US 10,289,184 B2
  • Filed: 10/07/2015
  • Issued: 05/14/2019
  • Est. Priority Date: 03/07/2008
  • Status: Active Grant
First Claim
Patent Images

1. A method for cognizant power management in a distributed computing system including a computing device, servers, power supply units and cooling units, comprising the steps of:

  • gathering, by the computing device, values for electrical power metrics from the servers and power supply units, the electrical power metrics characterizing electrical power usage in the computing environment;

    gathering, by the computing device, values for computational power metrics from operating systems and applications executed within the servers, the computational power metrics characterizing a distribution of a computational load among the servers;

    gathering, by the computing device, values for environmental metrics from environmental monitoring devices in the computing environment, the environmental metrics characterizing environmental conditions in the computing environment;

    generating, by the computing device, a predictive behavior power utilization model based on the values of the electrical power metrics, the computational power metrics and the environmental metrics for the computing system, the predictive behavior power utilization model effective to predict values of electrical power metrics and environmental metrics as a function of the values of the computational power metrics;

    distributing, by the computing device, an application load to the distributed computing system;

    gathering, by the computing device, values for the computational power metrics associated with the application load;

    applying, by the computing device, the predictive behavior power utilization model to predict values for the electrical power metrics and the environmental metrics under the application load;

    gathering, by the computing device, values of the actual electrical power metrics and the actual environmental metrics under the application load;

    updating, by the computing device, the predictive behavior power utilization model when the predicted values for the electrical power metrics and the environmental conditions fall outside defined tolerance limits for the actual electrical power metrics and environmental conditions; and

    reallocating, by the computing device, the application load by executing an optimization step whereby the application load is reallocated to minimize a value of total electrical power utilization by the servers, power supply units and cooling units as determined by values of the electrical power metrics and environmental metrics predicted by the updated dynamic model as a function of computational power metrics associated with the reallocated application load.

View all claims
  • 1 Assignment
Timeline View
Assignment View
    ×
    ×