A Machine Learning Model for Detection and Prediction of Cloud Quality of Service Violation

Cloud services connect user with cloud computing platform where services range from Infrastructure as a Service, Software as a Service and Platform as a Service. It is important for Cloud Service Provider to provide reliable cloud services which are fast in performance and to predict possible service violation before any issue emerges so then remedial action can be taken. In this paper, we therefore experiment with five different machine learning algorithms namely Support Vector Machine, Random Forest, Naive Bayes, Neural Network, and k-Nearest Neighbors for the detection and prediction of cloud quality of service violations in terms of response time and throughput. Experimental results show that the model created using SVM incorporated with 16 derived cloud quality of service violation rules has consistent accuracy of greater than 99%. With this machine learning model coupled with 16 decision rules, the Cloud Service Provider shall be able to know before hand, whether violation of services based on response time and throughput is occurring. When transactions tend to go beyond the threshold limits, system administrator shall be alerted to take necessary preventive measures to bring the system back to normal conditions. This shall reduce the chance for violation to occur, hence mitigating lose or costly penalty.