The management of data within Grids is a challenging problem. It requires providing easy access to distributed, heterogeneous data that may reside in different ‘administrative domains’ and may be represented by heterogeneous data formats, and/or have different semantic meaning. Since applications may access data from a variety of storage repositories, for example, file systems, database systems, Web sites, document management systems, scientific databases, and so on, there is a need to define a higher-level abstraction for data organization. This is generally referred to as a data collection. A data collection contains named entities that may in actuality be stored in. | 16 Virtualization services for Data Grids Reagan W. Moore and Chaitan Baru University of California San Diego California United States INTRODUCTION The management of data within Grids is a challenging problem. It requires providing easy access to distributed heterogeneous data that may reside in different administrative domains and may be represented by heterogeneous data formats and or have different semantic meaning. Since applications may access data from a variety of storage repositories for example file systems database systems Web sites document management systems scientific databases and so on there is a need to define a higher-level abstraction for data organization. This is generally referred to as a data collection. A data collection contains named entities that may in actuality be stored in any of the repositories mentioned above. Data collections can be organized hierarchically so that a collection may contain subcollections and so on. An important issue in building collections is naming. Typically the naming conventions are dependent upon the type of digital entities. While there are technologies to build uniform namespaces they are typically different for each type of digital entity. For example a Geographic Information System may use one approach to organize and name spatial data database systems may use another approach for structured Grid Computing - Making the Global Infrastructure a Reality. Edited by F. Berman A. Hey and G. Fox 2003 John Wiley Sons Ltd ISBN 0-470-85319-0 410 REAGAN W. MOORE AND CHAITAN BARU data and file systems may use yet another for files. Data Grids must provide the capability to assemble and further integrate such disparate data into coherent collections. This problem is further exacerbated owing to the fact that data collections can persist longer than their supporting software systems. Thus the organization of data collections may need to be preserved across multiple generations of supporting infrastructure. There is
