Predictive two-dimensional autoscaling

Embodiments may enable cloud based computing infrastructure to automatically scale in response to changing service demands. Auto-scaling may be enabled by automatically provisioning computing resources as they may be needed by hosted computing services. Historical utilization patterns may be tracked enabling the generation of models that may be employed to predict future computing resource requirements. The automatic scaling system may comprise one or more models that may be trainable using business rules that may be applied to determine to if and how computing resources are scaled. Further, business rules may be arranged to determine provisioning and scaling of computing resources based in part on the historical usage patterns of the computing services.