EEMD-ARIMA based cloud computing server load short-term prediction method

Abstract

The invention discloses an EEMD-ARIMA based cloud computing server load short-term prediction method. The EEMD-ARIMA based cloud computing server load short-term prediction method comprises the steps of firstly building a training set for load data; then carrying out EEMD decomposition for the load data to obtain multiple IMF components and remainder terms; then calculating an information entropy, a correlation coefficient and an energy factor of each IMF component and each remainder term, and further building an effective evaluation factor to select effective IMF components and effective remainder terms; and finally carrying out ARIMA prediction and summing for the effective components to obtain a final prediction result. Precise short-term prediction for the cloud computing server load is realized. The EEMD-ARIMA based cloud computing server load short-term prediction method is suitable for prediction of unsteady load of the cloud computing server, in particular, for the prediction of an unsteady state condition of the server load possibly caused by randomly burst user resource demands in a short time in a public cloud, is suitable for all the load types used by cloud computing and is not limited by a single load type.

Description

technical field [0001] The present invention relates to a short-term prediction method of cloud computing server load, and more specifically, relates to a short-term prediction method of cloud computing server load based on EEMD-ARIMA. Background technique [0002] Cloud computing gathers a large number of computing, storage, and network resources to form a large resource pool, and effectively cuts, allocates, and integrates resources through virtualization technology to achieve efficient use of resources. It provides an on-demand mode. Cloud computing users can lease resources on demand to reduce costs. Cloud platform providers can realize on-demand supply and dynamic management of virtual resources through reasonable scheduling to improve resource utilization. The characteristics of on-demand supply, low cost, and easy expansion make cloud computing widely used. However, due to the different application requirements of users and the heterogeneity of cloud platform resource...

AI Technical Summary

Problems solved by technology

Because the first type of method depends on the level of expert knowledge, the results are difficult to be explained theoretically, and they are rarely used in recent years; the second and third types of methods are widely used in forecasting, but most of them are based on relatively stationary time series data make long-term forecasts

In the cloud computing environment, especially the public cloud, the demand for user resources is random and sudden, which may cause the server load to appear non-stationary in a short period of time, and it is difficult to collect a large amount of sample data in a short period of time; artificial intelligence prediction methods are generally based on learning Training requires a large amount of sample data and is not suitable for short-term forecasting, and simple statistical forecasting models cannot obtain good forecasting results when dealing with non-stationary data

Images