An Architecture for Management of Large, Distributed, Scientific Data Using SQL/MED and XML
Mark Papiani, Jasmin Wason, Denis A. Nicole · ePrints Soton (University of Southampton) · 2000
We have developed a Web-based architecture and user interface for archiving and manipulating results of numerical simulations being generated by the UK Turbulence Consortium on the United Kingdom's new national scientific supercomputing resource. These simulations produce large datasets, requiring Web-based mechanisms for storage, searching and retrieval of simulation results in the hundreds of gigabytes range. We demonstrate that the new DATALINK type, defined in the draft SQL Management of External Data Standard, which facilitates database management of distributed external data, can help to overcome problems associated with limited bandwidth. We show that a database can meet the apparently divergent requirements of storing both the relatively small simulation result metadata, and the large result files, in a unified way, whilst maintaining database security, recovery and integrity. By managing data in this distributed way, the system allows post-processing of archived simulation results to be performed directly without the cost of having to rematerialise to files. This distribution also reduces access bottlenecks and processor loading. We also show that separating the user interface specification from the user interface processing can provide a number of advantages. We provide a tool to generate automatically a default user interface specification, in the form of an XML document, for a given database. The XML document can be customised to change the appearance of the interface. Our architecture can archive not only data in a distributed fashion, but also applications. These applications are loosely coupled to the datasets (in a many-to-many relationship) via XML defined interfaces. They provide reusable server-side post-processing operations such as data reduction and visualisation.