Petroleum engineering is a complicated subject which involves multidisciplinary teams of scientists, engineers, operators and managers to use real-time and historic data to perform modeling, interpretation, analysis and decision-making activities. During this process, some very crucial tasks come from how to effectively collect and manage the acquired datasets, how to efficiently provide data services, and how to control and assess data quality. In addition, heterogeneous sources and various types of the datasets make the process more complicated. To achieve effective data engineering, metadata is discussed in this paper. Furthermore, a comprehensive, extensible and operational metadata model is presented. The model is discussed from different perspectives. Conceptually, it is a metadata framework with four levels of abstraction. Logically, it characterizes a three-dimensional geological data map that interconnects geological entities on the plane and mounts data resources as sub-trees on the other dimension. Physically, it adopts a federated topology and supports distributed storage and management by virtual nodes mapping. Finally, how the metadata model works in data governance, collection, management, and service is illustrated.