Research and Analysis of Unsteady-State Processes in Cloud Systems

Abstract

This study is related to the research and analysis of unsteady-state processes in cloud systems. The research objective of this study is to provide the capability for prediction of effects of unsteady-state processes on operational metrics of cloud systems. The research subject is queuing systems used as models of cloud systems. The scope of research is unsteady-state processes and cloud systems operation specifics. To achieve the stated objective, this study solves several problems, including the classification of unsteady-state processes; development of a cloud system model and analysis of its properties when exposed to unsteady-state load processes and others. The theoretical background of the research includes scientific papers of domestic and foreign scientists in the field of cloud technology, queuing theory, queuing systems and networks theory. This study proposes a classification of unsteady-state processes according to the driving distributions and process nature. A composition method for unsteady-state distributions is characterized. These results allow predicting loads on cloud systems with load processes consisting of several streams with different intensities and frequencies. The concluded research allowed to analyse the effects of unsteady-state processes and their composition on the operation specifics of queue systems used as models of cloud systems. The results obtained allowed to determine the efficiency of cloud systems based on the parameters of unsteady-state load processes, which can positively affect cloud system design and management.The main results and conclusions of this research are presented at the end of this paper. It is noted that the obtained analytical dependencies, the developed cloud system simulation model and can be used in cloud system design and management, in particular, in automatic scaling algorithms for cloud systems and container-based cluster orchestration systems.