METHOD FOR IDENTIFICATION INFORMATION SERVICES IN VIRTUAL MACHINES OPERATING IN A DISTRIBUTED TECHNICAL SYSTEM

The problem of the availability of computing resources in distributed technical systems is considered, which arises in connection with the imbalance of servers in the absence of information about the type of information service in virtual machines when implementing the "Infrastructure as a service" cloud computing service model. The proposed method of identifying services in virtual machines includes several stages. At the first stage, the processor and RAM load is monitored for current information services in order to obtain time series of values obtained from all existing counters, and then using the chisquare criterion, the dimensionality of the task of their analysis is reduced. At the second stage, after testing hypotheses about the distribution of random variables using the Kolmogorov agreement criteria and the Shapiro-Wilk test, the parameters of the distribution density function are calculated, which are selected as classification features. At the third stage, using the random forest method embedded in the scikitlearn machine learning library, the classifier is trained and the classification results are compared with real data. The conducted series of experiments showed a fairly high value of classification accuracy in terms of Accuracy, on average equal to 0.78, and ensuring uniform loading of servers using the proposed method of identifying information services compared with the standard load balancing algorithm in the Brest cloud platform.