Back to results
US 11,082,301 B2Granted
Forecasting computer resources demand
AssigneeInternational Business Machines Corporation
InventorsAaron K. Baughman, Richard J. Bogdany, Richard A. Locke, Brian M. O'Connell
Life of this patent
Estimating the term…
Abstract
An approach for forecasting demand. The approach includes a method that includes receiving one or more variables associated with an event. The method further includes generating, by at least one computing device, a model to forecast future demand based on the one or more variables. The method further includes determining, by the at least one computing device, a load to provision one or more servers to meet the future demand. The load is based on the model.
Claims18
A method comprising:
- generating, by at least one computing device, a model to forecast future demand based on one or more variables;
- determining, by the at least one computing device, a load to provision one or more cloud servers to meet the forecast future demand, wherein the load is based on the model;
- simulating, by the at least one computing device, based on a trained model, web data traffic demands associated with a participant in an event, a portion of the event, and a playoff/runoff scenario;
- physically provisioning the one or more cloud servers with the determined load based on the trained model which includes a forecasted future demand with a threshold level of error to increase forecasting accuracy during an 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.
3 more elements
A method comprising:
- generating, by at least one computing device, a model to forecast future demand based on one or more variables;
- determining, by the at least one computing device, a load to provision one or more cloud servers to meet the forecast 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 future demand with a threshold level of error to increase forecasting accuracy during an event;
- 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;
- generating at least one event in a future timeline;
- generating at least one predicted burst event based on a historical timeline of a previous event;
- generating a predicted spike in demand based on the at least one predicted burst event using the historical timeline;
- generating a spike in demand based on a current timeline;
- adding overlap to a forecast curve which is a difference between the predicted spike in demand based on historical information and the spike in demand based on the current timeline;
- generating a spike adjusted forecast based on the generated predicted spike in demand, the generated spike in demand, and the overlap;
- and
- determining a load based on the generated spike adjusted forecast.
9 more elements
A method comprising:
- generating, by at least one computing device, a model to forecast future demand based on one or more variables;
- determining, by the at least one computing device, a load to provision one or more cloud servers to meet the forecast 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 future demand with a threshold level of error to increase forecasting accuracy during an event;
- 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;
- generating a linear regression model that includes the one or more variables associated with event-related features;
- determining a weight for each of the one or more variables based on a predicted web traffic;
- determining an updated predicted web traffic based on the linear regression model with the determined weight for each of the one or more variables;
- and
- allocating computing resources to meet the updated predicted web traffic.
6 more elements
Dependencies are inferred from the claim text.