Application Service Behavior Prediction Model Over Inter-cloud Environment

No Thumbnail Available



Journal Title

Journal ISSN

Volume Title


Addis Ababa University


Cloud computing is a computing model that delivers different services to its users through Internet. These services include storage, databases, networking, analytics and software. Nevertheless, the delivery of these services towards the users will be difficult, if resources on the cloud are overloaded due to increased workloads. To overcome this situation an environment called Inter-cloud environment is designed. This environment is designed by forming cloud of clouds where each cloud would use the computational, storage or any kind of infrastructural resource of other clouds. However, the aggregation of diversified computing systems in the Inter-cloud environment poses difficult problems in effective delivery of application services and resource provisioning. These problems arise because of the magnitudes and uncertainties of Inter-cloud components (workload, compute servers, services). This research aims to study the Inter-cloud environment along with the behaviors of the application services and to propose a prediction model that assists the environment with knowledge to future resource surge of each service. The application service behaviors prediction model will be used to predict the CPU utilization of Inter-cloud services. The prediction model was developed by the most widely used machine learning method, Artificial Neural Network (ANN). Among the Artificial Network Algorithms; Multilayer Perceptron Neural Network (MLP) is used to approximate any linear or non-linear function. MLP method is employed to develop application service behavior prediction model. The Inter-cloud environment is simulated using FederatedCloudsim framework. Materna workload traces and Bitbrain workload traces are used to generate random resource workload traces from the FederatedCloudsim framework. The generated resource workload traces have been used to analyze the problem, to train and test the proposed prediction model. Four experiments were designed to build the application service behavior prediction models using generated resource workload data of Materna workload and Bitbrain workload traces from Inter-cloud environment. From the evaluation of the prediction model two factors that could affect the accuracy of the predicted results are pointed out. In this work the Coefficient of Determination and Mean Squared Error metrics are used to analyze the accuracy of the predictor model.



Inter-Cloud Environment, Application Service Behavior, ANN, MLP, Machine Learning, Prediction Model