US 10,476,949 B2 · Predictive autoscaling in computing systems

Abstract

Techniques of predictive autoscaling in distributed computing systems are disclosed herein. In one embodiment, a method includes receiving data representing consumption of a computing resource by an application executing on one or more servers in the distributed computing system. The method also includes processing the received data into a time series having multiple resource consumption values by the application with corresponding time stamps and decomposing the time series into a regular component and an irregular component. The method further includes generating a predicted consumption value of the computing resource by the application at a future time point according to the trend, cyclic pattern, or seasonal pattern of the regular component of the time series and causing immediate adjustment of an amount of the computing resource provisioned in the distributed computing system for the application according to the generated predicted consumption value.

Claims20

Dependent claims

A method performed by a computing device in a distributed computing system having a plurality of servers interconnected by a computer network, the method comprising:

receiving data representing consumption values of a computing resource by multiple instances of an application executing on respective multiple servers in the distributed computing system, the computing resource includes a processor capacity, network capacity, or storage capacity;
processing the received data into a time series having multiple resource consumption values by the application via aggregating the consumption values of the computing resource by the multiple instances of the application on the multiple servers in accordance with corresponding time stamps;
decomposing the time series into a regular component and an irregular component, the regular component representing a portion of the time series having one or more of a trend, a cyclic pattern, or a seasonal pattern while the irregular component representing a random portion of the time series;
generating a predicted consumption value of the computing resource by the application at a future time point according to the trend, cyclic pattern, or seasonal pattern of the regular component of the time series;
and
causing immediate adjustment of an amount of the computing resource provisioned in the distributed computing system for the application according to the generated predicted consumption value.

3 more elements

A method performed by a computing device in a distributed computing system having a plurality of servers interconnected by a computer network, the method comprising:

receiving, via the computer network, data representing a predicted consumption value of a computing resource used by an application at a future time point in the distributed computing system, the application having multiple instances executing on respective multiple servers, the computing resource includes one or more of a processor capacity, network capacity, or storage capacity, wherein the predicted consumption value is set based on a trend, a cyclic pattern, or a seasonal pattern of historical consumption values of the computing resource used by the multiple instances of the application in the distributed computing system by aggregating consumption values of the computing resource by the multiple instances of the application on the respective multiple servers in accordance with corresponding time stamps;
and
in response to the received data, adjusting an amount of the computing resource provisioned in the distributed computing system for the application according to the received predicted consumption value when an amount of the computing resource currently consumed by the application does not exceed a preset overutilization threshold or a underutilization threshold.

A computing device in a distributed computing system having a plurality of servers interconnected by a computer network, the computing device comprising:

a processor;
and
a memory containing instructions executable by the processor to cause the processor to:
receive data representing consumption values of a computing resource by multiple instances of an application executing on respective multiple servers in the distributed computing system, the computing resource includes a processor capacity, network capacity, or storage capacity;
generate a time series based on the received data by aggregating the consumption values of the computing resource by the multiple instances of the application on the multiple servers in accordance, the time series having multiple temporal resource consumption values;
derive a regular component of the time series from the generated time series, the regular component representing a portion of the time series having one or more of a trend, a cyclic pattern, or a seasonal pattern;
predict a future consumption value of the computing resource by the application at a future time point according to the trend, cyclic pattern, or seasonal pattern of the regular component of the time series;
and
adjust an amount of the computing resource provisioned in the distributed computing system for facilitating execution of the application according to the generated future consumption value.

6 more elements

Dependencies are inferred from the claim text.

Predictive autoscaling in computing systems

Life of this patent

Predictive autoscaling in computing systems

Life of this patent

Abstract

Claims20