Інтелектуальна система управління кластерною системою на базі балансування навантаження

Abstract

This master diploma work is devoted to the development of automated load balancing program on the clustered system. In this work, modern methods, technical and existing software for predicting the load on the cluster system were considered depending on the predicted output. Based on the conducted analysis as for forecasting a 2 step method has been developed that’s using logistic regression via Apache Mahout library and fuzzy ruleset provided by domain experts. Developed algorithmic and software basis for a prediction system using C# programming language. Investigation of developed method efficiency and software testing proved correctness of retrieved results. Proposed automated cluster load forecasting system allows to increase efficiency of computing resources utilization in cloud environments and implement better system reacting to abruptly load changes.