This thesis introduces a holistic model of predictive autoscaling for multilayered cloud deployments. The multilayered aspect is reflected in treating autoscaling as a hierarchical process where scaling on the application level acts as a driver for scaling on the level of VM clusters. The following processes constitute the predictive aspect: load forecasting, performance modeling, application topology analysis, and schedule derivation. Each of them is studied in the thesis to allow for justified design decisions when composing a customized predictive autoscaling policy.
«
This thesis introduces a holistic model of predictive autoscaling for multilayered cloud deployments. The multilayered aspect is reflected in treating autoscaling as a hierarchical process where scaling on the application level acts as a driver for scaling on the level of VM clusters. The following processes constitute the predictive aspect: load forecasting, performance modeling, application topology analysis, and schedule derivation. Each of them is studied in the thesis to allow for justified...
»