New challenges for large-scale data management and sharing have arisen because of the information evolution from massive data to "big data". In this paper, we propose the concept of virtual dataspaces and domain scientific data cloud (DSDC) to address the challenges of management, reusability, and service of scientific data in scientific domain. The application of DSDC is also introduced, which is used in the materials science domain for e-Science applications. Virtual dataspaces is then defined, which is used to integrate and organize both structured and unstructured scientific data in cloud environment. To build the virtual dataspaces, a logical resources model based on ontology is discussed. And then a new mapping and evolutionary model of virtual dataspaces is described in a "pay-as-you-go" fashion. As such, we introduce a "semi-automatic evolution" method. A scientific data cloud application model is also proposed for defining service template. Finally, the virtual dataspaces and DSDC are verified by "material service safety evaluation", and the results show that our approach is efficient in scientific data management and suitable for data-intensive application.Management System (DBMS) could only store and query structured data, but could not deal with all kinds of data. How should scientists manage various "big data" efficiently? Taking material data, for example, there are several systems deployed in different servers to manage material data with relational Relational Database Management System (RDBMS) that stored data in a scientific research group. The databases are probably distributed in different servers. Scientists probably could not obtain useful information from complex and massive material data by using relational RDBMS that are lack of semantic information. Most semistructured and unstructured data are stored in local file system, such as MatML files, text files, equilibrium diagram, video, and so forth. These data normally involve more information for materials of scientists. Thus, how should scientists obtain useful information from these data?We discuss the requirements of scientists by several questions. Since the continuous increase of quantity and complexity of data, how can one guarantee the reuse of scientific data on the Web? How can one provide the scientists exact information by mining semantic relationship among scientific data? For instance, if a scientist would like to inquire data associated with "pitting corrosion of carbon steel in soil environment", the keywords of which could be stored in relational RDBMS, paper, text files, and XML. However, it is difficult to achieve by structured query and keywords searching in distributed and heterogeneous "big data" environment. In this situation, how can one make sure scientists could obtain exact information?Application services for scientists are the ultimate goal of managing scientific "big data" effectively. How can one implement register, reuse, combination, and interoperability for services to publish and use them conveniently and f...