The operation of a cloud-based IT system (system for short) is time-consuming and error-prone due to the system scale, heterogeneity and configuration dependency. Although administrators can manage their systems with various configuration management tools, a plenty of knowledge spanning various domains is necessary. To alleviate this situation, we present a model-driven service MORE (Model-driven Operation seRvicE) to automate the initial deployment and the dynamic configuration of a system. Firstly, a model is proposed to specify the high-level view of a system in the form of a desired deployment topology. Then the topology model is transformed into executable code automatically, bridging the gap between high-level abstractions and low-level details. With those executable code as input, a runtime framework is designed based on a transaction-based self-configuration protocol to achieve automation and configuration consistency. Finally, we evaluate the service abilities (including modeling system topology, automating system provisioning, performing runtime reconfiguration) with a case study.