USD - a database management system for scientific research

Abstract

USD (UnstructuredScientific Data) is a database system developed at Lawrence Livermore National Laboratory (LLNL) that provides database capabilities required when doing scientific research. USD is implemented in a version of EMACS Lisp that has been extended to include a relational database management system and graphical user interface primitives. It is currently being used by several dhlerent scientific research efforts at LLNL.Current data models require that data be manipulated in terms of a predcfmed structure. Many scientific database applications have stmctured dat~but also have unstructured data that is not easily organized. The need for unstructured data arises in applications where the organization is unknown or continually changing .The existence of this type of data appears to be a driving force behind the development of object oriented data models and the extended relational data model. In essence, the strategy employed by these models is make complex structure easier to deal with.However, structure must be pre-specified and adhered to. That is, prior to data entry, the data organization must be determined and all data must be expressed in terms of that stmcture. From this perspective, these data models arc not much different from the relational data model (or from the network and hierarchical data models). The crucial aspect of this work is in accepting the unstructured view of this data.Another aspect of scientific research that USD addresses is the widespread use of pre-existing data analysis tools. Both inhouse as well as commercial products are in use. In general, these tools are used in both batch and interactive modes. USD provides a mechanism whereby it is easy to construct an interface for anew tool without having to modify the tool itself. The tool can then be run as a foreign process to USD and be operated in either batch or interactive modes.The video presentation shows three examples of USD being used to analyze seismic data from an underground nuclear explosions.The first example introduces the basics required to use USD.USD employs semantic networks as a user data model. A semantic network is a labclled directed graph wherein nodes represent objects and links represent the relationships between those objects.There exists an obvious, well defined mapping from a relational representation to a semantic network representation. Semantic networks can also be used to represent unstructured data. Data retrieval is illustrated by specifying a semantic network pattern (i.e. a semantic network with labelled links but no objects at the nodes) and directing USD to mutch that pattern. The resulting data is displayed as a sequence of semantic networks. Data is also displayed in tabular form. Fkmlly, unstructured data is appended through via the graphical user interface.The second example illustrates the capability to use a foreign process in order to analyze data. FirsL a pattern is constructed by merging two existing patterns and then modifying it to match just the data required. A procedure patter...

ViewFinder: an object browser

“…Facts are named binary relationships between data values, so the data may be regarded as a semantic network of values. This loosely-structured model of databases has been adopted by USD [22], a system that provides database capabilities often required in scientific research. The main retrieval and browsing paradigm of its graphical user interface is the matching of patterns constructed by users.…”

A Hierarchy Navigation Framework: Supporting Scalable Interactive Exploration over Large Databases

“…Database support systems for visualization systems such as Tioga [22], USD [14], GODVIA [17] and IDEA [21] represent work closely related to ours in the sense that all of them work towards making visualization systems run over large persistent data. However, many of the detailed techniques used are rather different in each of these systems.…”

“…However, many of the detailed techniques used are rather different in each of these systems. USD [14] proposes a semantic net model to store and retrieve unstructured data. I Tioga [22] implements a multiple browser architecture for visual queries.…”