Predicting QoS of virtual machines via Bayesian network with XGboost-induced classes

Quality of Service (QoS) of virtual machines (VMs) is guaranteed by the Service Level Agreements (SLAs) signed between users and service providers during the renting of VMs. A typical idea to ensure the SLAs being reached is to predict the QoS of VMs accurately and then take the appropriate measures according to the prediction results timely. However, the QoS is affected by multiple VM-related features, among which the uncertain and non-linear relationships are challenging to represent and analyze. Thus, in this paper, we construct a class parameter augmented Bayesian Network (CBN) to overcome the difficulties and then predict the QoS of VMs accurately. Specifically, we first cluster multiple VM-related features based on the Euclidean distance, and then use XGboost to classify the different VM configurations within each cluster. Then, we construct the CBN based on the classification results as well as the corresponding QoS values. Consequently, we predict the QoS of VMs via the variable elimination (VE) with CBN. Experimental results show the efficiency and effectiveness of our proposed method on predicting the QoS of VMs.