Memory usage determination techniques

Abstract

Embodiments provide techniques for estimating seasonal indices for multiple periods. Some embodiments can receive a signal comprising a plurality of measures sampled over a span of time from an environment in which one or more processes are being executed. Some embodiments may then extract a seasonal effector and a de-seasonalized component from the signal. Next, some embodiments can apply one or more spline functions to the seasonal effector to generate a first model. Some embodiments may then apply a linear regression technique to the de-seasonalized component to generate a second model. Some embodiments may then initiate actions associated with the code. Some embodiments may then generate a forecast of the signal based on the first model and the second model. Next, some embodiments may initiate, based at least in part on the forecast, one or more actions associated with the environment.

What is claimed is: 
     
         1 . A computer-implemented method comprising:
 receiving, by one or more computer systems, a signal comprising a plurality of measures sampled over a span of time from a cloud computing environment in which one or more processes are being executed;   extracting a first component of the time-series measurement having seasonal factors and a second component of the time-series measurement being de-seasonalized;   applying one or more spline functions to the first component to generate a first model, wherein applying the one or more spline functions to the first component comprises assigning first weights to samples of the first component, each first weight being determined based at least in part on a difference between each sample and an expected value of the sample, and wherein the first weight for a sample increases when the difference decreases;   applying a linear regression technique to the second component to generate a second model;   generating a forecast of the signal based at least in part on the first model and the second model; and   initiating, based at least in part on the forecast, one or more actions associated with the cloud computing environment.   
     
     
         2 . The method of  claim 1 ,
 wherein the span of time spans a plurality of cycles of a period having a particular length;   wherein the period is divided into a plurality of regular intervals; and   wherein extracting the first component and the second component comprises:
 for each of the plurality of regular intervals, determining an average measure of the interval; 
 for each of the plurality of cycles, determining an average measure of the cycle; 
 determining a set of seasonal factors by, for each of the plurality of regular intervals, determining a season factor for the interval by comparing the average measure of the interval with the average measure of the cycle; 
 applying a spline function to the set of seasonal factors to obtain the first component; and 
 de-seasonalizing the time-series measurement based at least in part on the first component to obtain the second component. 
   
     
     
         3 . The method of  claim 1 , wherein the plurality of measures are sampled at irregular intervals over the span of time. 
     
     
         4 . The method of  claim 3 , wherein the irregular intervals at which the plurality of measures are sampled and the plurality of measures exhibit a dependency relationship, wherein the linear regression technique is a robust linear regression technique, and wherein the robust linear regression technique is applied to the second component to compensate for the dependency relationship. 
     
     
         5 . The method of  claim 4 , wherein applying the robust linear regression technique to the second component comprises, for each of the plurality of measures, assigning a weight to the measure based at least in part on a length of an irregular interval associated with the measure. 
     
     
         6 . The method of  claim 5 , wherein applying the robust linear regression technique to the second component further comprises, for each of the plurality of measures, trimming the measure if the length of the irregular interval associated with the measure does not exceed a threshold length. 
     
     
         7 . The method of  claim 5 , wherein applying the robust linear regression technique to the second component comprises, for each of the plurality of measures:
 predicting an expected measure that corresponds to the measure; and   assigning a weight to the measure based at least in part on a deviation between the expected measure and the measure.   
     
     
         8 . The method of  claim 1 , wherein the signal is heteroscedastic, and wherein the linear regression technique is applied to the second component to account for the heteroscedasticity of the signal. 
     
     
         9 . The method of  claim 1 , wherein the signal corresponds to a usage of a heap, over the span of time, of the cloud computing environment. 
     
     
         10 . The method of  claim 1 , wherein the one or more actions comprises providing additional resources to the cloud computing environment. 
     
     
         11 . A system comprising:
 one or more processors; and   a memory accessible to the one or more processors, the memory storing one or more instructions that, upon execution by the one or more processors, causes the one or more processors to:
 receive a signal comprising a plurality of measures sampled over a span of time from a cloud computing environment in which one or more processes are being executed; 
 extract a first component of the time-series measurement having seasonal factors and a second component of the time-series measurement being de-seasonalized; 
 apply one or more spline functions to the first component to generate a first model, wherein applying the one or more spline functions to the first component comprises assigning first weights to samples of the first component, each first weight being determined based at least in part on a difference between each sample and an expected value of the sample, and wherein the first weight for a sample increases when the difference decreases; 
 apply a linear regression technique to the second component to generate a second model; 
 generate a forecast of the signal based at least in part on the first model and the second model; and 
 initiate, based at least in part on the forecast, one or more actions associated with the cloud computing environment. 
   
     
     
         12 . The system of  claim 11 ,
 wherein the span of time spans a plurality of cycles of a period having a particular length;   wherein the period is divided into a plurality of regular intervals; and   wherein extracting the first component and the second component comprises:
 for each of the plurality of regular intervals, determining an average measure of the interval; 
 for each of the plurality of cycles, determining an average measure of the cycle; 
 determining a set of seasonal factors by, for each of the plurality of regular intervals, determining a season factor for the interval by comparing the average measure of the interval with the average measure of the cycle; 
 applying a spline function to the set of seasonal factors to obtain the first component; and 
 de-seasonalizing the time-series measurement based at least in part on the first component to obtain the second component. 
   
     
     
         13 . The system of  claim 11 , wherein the plurality of measures are sampled at irregular intervals over the span of time. 
     
     
         14 . The system of  claim 13 , wherein the irregular intervals at which the plurality of measures are sampled exhibit a dependency relationship, wherein the linear regression technique is a robust linear regression technique configured to be applied to the second component to compensate for the dependency relationship. 
     
     
         15 . The system of  claim 14 , wherein applying the robust linear regression technique to the second component further comprises, for each of the plurality of measures, trimming the measure if the length of the irregular interval associated with the measure does not exceed a threshold length. 
     
     
         16 . The system of  claim 15 , wherein applying the robust linear regression technique to the second component further comprises, for each of the plurality of measures, trimming the measure if the length of the irregular interval associated with the measure does not exceed a threshold length. 
     
     
         17 . The system of  claim 11 , wherein the signal is heteroscedastic, and wherein the linear regression technique is applied to the second component to account for the heteroscedasticity of the signal. 
     
     
         18 . A non-transitory computer-readable medium storing one or more instructions that, when executed by a processor, cause the processor to perform operations comprising:
 receiving, by one or more computer systems, a signal comprising a plurality of measures sampled over a span of time from a cloud computing environment in which one or more processes are being executed;   extracting a first component of the time-series measurement having seasonal factors and a second component of the time-series measurement being de-seasonalized;   applying one or more spline functions to the first component to generate a first model, wherein applying the one or more spline functions to the first component comprises assigning first weights to samples of the first component, each first weight being determined based at least in part on a difference between each sample and an expected value of the sample, and wherein the first weight for a sample increases when the difference decreases;   applying a linear regression technique to the second component;   generating a forecast of the signal based at least in part on the first model and the second model, the second model being generated by applying the linear regression; and   initiating, based at least in part on the forecast, one or more actions associated with the cloud computing environment.   
     
     
         19 . The non-transitory computer-readable medium of  claim 18 , wherein
 applying the linear regression includes assigning second weights to samples of the second component, each second weight being determined based at least in part on a duration of an interval from which a sample is taken, and wherein the second weight for the sample increases when the interval increases.   
     
     
         20 . The non-transitory computer-readable medium of  claim 18 , wherein the forecast of the signal comprises a growth rate forecast of the signal, and wherein an amount of allocated memory resource is determined based at least in part on the growth rate forecast.