For example,Бобцов

ESTIMATION OF PERIODS OF NONSTATIONATY PROCESSES IN CLOUD SYSTEMS

Annotation

The problem of automated evaluation of load process period in cloud systems is considered. A common approach to load simulation for such systems implies the stationary nature of incoming requests flow, which results in inadequate matching of the models to the objects under investigation. To automatically construct an adequate analytical model, it is necessary to estimate the length of the loading process period from measured data. A numerical method of automated period estimation for any periodic functions is proposed. The method allows real-time processing of measured values, restoring missing data by interpolation of known neighboring values. Pearson correlation coefficient is chosen as a criterion to calculate functions similarity. It’s proposed to store arrays of sums of mean values and standard deviation values in the main memory to improve the method’s performance. Computational complexity of the method is reduced to polynomial, which means that the method could be implemented even in the embedded systems. Period estimation accuracy is demonstrated by presented results of numerous experiments with ideal, randomly noised, and real signals.

Keywords

Articles in current issue