The Research And Implementation Of Federated Data Service System Based On Coordination Of Multi-domain

In recent years, data volume in all kinds of large-scale scientific researches, information services, multimedia applications, etc presents explosive increasing. And the needs for integrated access, transfer, management of geographically distributed mass datasets are also increasing. The data grid technology is an effective solution to this problem and is also a research hotspot in computer field in the world. Since 1990s, the USA and some European countries have had some important achievements in fundamental theories and system development of grid computing. Relative researches in grid computing are now carried on actively in our country.Our project has studied the key technologies of data grid, and has designed and implemented a general data grid middleware called GridDaEn, which can achieve uniform access, transfer and management of distributed, heterogeneous mass datasets and provide a high-performance platform for data-intensive applications and researches over wide area networks.This paper focuses on the key technologies of federated data service in data grid, and describes the design and implementation of FDSS, the core of GridDaEn data grid system, and DRB, the basic data service provider of FDSS. DRB can seamlessly integrate distributed and heterogeneous storage resources in wide area networks into the GridDaEn data grid by providing scalable uniform access interface and supporting multiple access modes. DRB can provide a uniform user view of distributed, heterogeneous mass storage resources by virtual organization and management of storage resources, and thus shields the distribution and heterogeneity of resources and achieves completely transparent access. A single DRB provides basic local-domain data service. Multiple DRBs form the FDSS through federated approach and provide federated data service based on coordination of multi-domain. DRB adopts perfect cache management, concurrency control mechanism of access to shared files based on read-write lock and security mechanism based on Globus GSI to ensure FDSS's providing high-performance, consistent, secure federated data service. FDSS also provides a Windows Explorer-like client program, a set of command-line tools and standard uniform access API and SDK. By using these client software, users can fully utilize the federated data service provided by FDSS and can conveniently construct other grid applications.The key of achieving the federated data service in FDSS is the uniform and efficient coordination of DRBs. In this paper, an algorithm called Operation Decomposition and Time Division is presented to achieve the coordination of multiple DRBs. The test results indicate that the algorithm can achieve uniform and high-efficient coordination of multiple DRBs in FDSS.