http://www6.informatik.uni-erlangen.de:1200/Research/sdm.html (Einblicke ins Internet, 10/1995)
Horst L├╝hrsen
Scientific Database Management
Description
Scientific applications (e.g. climate research, manufacturing process control,
marketing research) require data management support that differs significantly
from traditional database applications like finance and administration. Even
so-called non-standard database systems do not adequately address issues
prevailing in scientific areas, for example time series analysis or multi-
dimensional data modelling. Some substantial research has been carried out in
the spatial and time domain, but mostly independently from one another and not
focussed on the specific demands of scientific computing. In particular, little
attention has been given to data modelling and administration tasks found in
scientific computing in conjunction with physical database design and query
optimization.
The research at IMMD VI concentrates on multi-dimensional arrays as a generic
data model for scientific applications. Typically, some measured values (e.g.
temperature values) can be described by a multitude of dimensions (e.g. space,
time). Queries against these data make use of categorizations that can be
defined for the different dimensions (e.g. days-weeks-months-years in the
time domain). Our approach is to store data according to the dimensional
categorizations in so-called packages and support query processing by
materializing summary data (e.g. mean temperature for different month) for
selected packages. Often, queries can be answered using the summary data
instead of the raw data. The package model also allows for an integration of
hierarchical storage management into query optimization.
More information
As this is an ongoing research effort, more information should be requested
by sending an e-mail or fax to Thomas Ruf.
Contact
Thomas Ruf
ruf@informatik.uni-erlangen.de