Forecasting computer resources demand
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.
3 Assignments
0 Petitions
Accused Products
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.
-
Citations
20 Claims
-
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, and the 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 Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15)
-
-
16. A computer program product for predicting demand in a cloud environment, the computer program product comprising a computer usable storage medium having program code embodied in the storage medium, the program code being readable/executable by a computing device to:
-
receive one or more variables associated with an event; generate a model to forecast future demand based on the one or more variables being associated with spikes of demand over a period of time; determine a load based on the model, wherein the load is used to provision one or more cloud servers to meet the future demand; physically provision the one or more cloud servers with the determined load based on a trained model which includes a burst event to increase forecasting accuracy during the event; and automatically allocate additional cloud computing resources to meet the future demand ahead of a time of the event to prevent an unsustainable volume of web traffic, wherein the model is trained so that the future demand is forecasted with a threshold level of error, and the training of the model includes generating the 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 with a regularized stochastic gradient descent, the number of coefficients being equal to the magnitude value of the demand factor feature vector. - View Dependent Claims (17, 18, 19)
-
-
20. A system comprising:
-
a CPU, a computer readable memory and a computer readable storage medium; program instructions to receive one or more variables associated with an event, wherein the one or variables are associated with data demand for playoff predictions; program instructions to generate a model to forecast future demand based on the one or more variables being associated with spikes of demand over a period of time, wherein the model uses a demand factor feature vector; and program instructions to determine a load based on the model, wherein the load is used to provision one or more cloud servers to meet the future demand, wherein the program instructions are stored on the computer readable storage medium for execution by the CPU via the computer readable memory, a web traffic associated with the playoff predictions is based on projecting web traffic for a predicted playoff based on analyzing different participants and their scoring capabilities within the event, physically provision the one or more cloud servers with the determined load based on the generated model which includes a forecasted future demand to increase forecasting accuracy during the event, automatically allocate additional cloud computing resources to meet the future demand ahead of a time of the event to prevent an unsustainable volume of the web traffic, and the additional cloud computing resources comprise the one or more cloud servers, at least one processor, and at least one memory device.
-
Specification