×

Forecasting computer resources demand

  • US 10,361,924 B2
  • Filed: 04/04/2014
  • Issued: 07/23/2019
  • Est. Priority Date: 04/04/2014
  • Status: Active Grant
First Claim
Patent Images

1. A method comprising:

  • receiving one or more variables associated with an event;

    generating, by at least one computing device, a model to forecast future demand based on the one or more variables;

    determining, by the at least one computing device, a load to provision one or more cloud servers to meet the future demand, wherein the load is based on the model;

    physically provisioning the one or more cloud servers with the determined load based on a trained model which includes a forecasted further demand with a threshold level of error to increase forecasting accuracy during the event; and

    automatically allocating additional cloud computing resources to meet the forecasted future demand ahead of a time of the event to prevent an unsustainable volume of web traffic,wherein the one or more variables are associated with the web traffic associated with at least one of proximity to lead or playoff predictions,the proximity to lead is based on projecting the web traffic associated with particular participants being within a threshold level of being in first place in the event, andthe model is trained so that the future demand is forecasted with the threshold level of error, and the training of the model includes generating a burst event based on a historical training set and an online training set, and the burst event is used to compute a demand factor feature vector which is used to train the model using a number of coefficients equal to the magnitude value of the demand factor feature vector.

View all claims
  • 3 Assignments
Timeline View
Assignment View
    ×
    ×