A database system for large, dynamically expanding three dimensional geomodels

Georg Semmler and Jan Gietzel and Helmut Schaeben. ( 2016 )
in: 2016 RING Meeting, ASGA

Abstract

In the past few years there was a steadily rising amount of three dimensional geological sub- surface models produced. The geometric variability of those models reaches from complex meshes consisting of triangles, tetrahedrons, pyramids or prisms to very large regular or irregular grids as well as combinations inside one geometry [Botella et al., 2015]. Commonly the choice of the geometric format depends on the application or the subsequent workflow. The cooperative work on constructing and modifying those models makes a central storage essential. Database systems storing and versioning geomodels have been developed in the last decade (e.g. GST, PostGIS, Rasdaman). Furthermore, this makes it easy to share or present the produced models over a network or the Internet. The increasing size and amount of produced models requires an efficient strategy for storing and accessing the primitives of a geometry. For large, high resolution geomodels, subselecting an area in a memory efficient way is a fundamental operation [Gietzel, 2011]. It is possible to provide those operations using relational database systems. Storing and accessing the geometry primitives in table records is becoming computational expensive for very large geomodels. The reason for this is that beside an overhead of the record access, adjacent information of a geometry primitives is not used by the database system to optimize the data layout on hard disk for spatial queries. Using this information it is possible to only access geometry primitives if they are actually required to process the query. In this paper we introduce a new technology to store the geometrical primitives in the file system to provide sub selection queries efficient, while retaining most of the advantages known from systems based on relational database systems. This system stores data in two different places. Metadata like name, size, author, used software for model construction, version, raw data connection and other data provenance information are stored in a relational database system. The large geometry data is stored outside of the relational database. This can be either a file based system or depending on custom requirements somewhere else e.g. as binary large object(BLOB) in a traditional database system, or on any other network storage device. The geometry is divided into spatially grouped tiles. Those tiles are the basic unit of storing the geometry. This combines the advantages of storing metadata in traditional environment and large geometry data sets in an optimized way for commonly used geometrical operations.

Download / Links

BibTeX Reference

@INPROCEEDINGS{,
    author = { Semmler, Georg and Gietzel, Jan and Schaeben, Helmut },
     title = { A database system for large, dynamically expanding three dimensional geomodels },
 booktitle = { 2016 RING Meeting },
      year = { 2016 },
 publisher = { ASGA },
  abstract = { In the past few years there was a steadily rising amount of three dimensional geological sub-
surface models produced. The geometric variability of those models reaches from complex meshes
consisting of triangles, tetrahedrons, pyramids or prisms to very large regular or irregular grids
as well as combinations inside one geometry [Botella et al., 2015]. Commonly the choice of the
geometric format depends on the application or the subsequent workflow.
The cooperative work on constructing and modifying those models makes a central storage
essential. Database systems storing and versioning geomodels have been developed in the last
decade (e.g. GST, PostGIS, Rasdaman). Furthermore, this makes it easy to share or present the
produced models over a network or the Internet.
The increasing size and amount of produced models requires an efficient strategy for storing
and accessing the primitives of a geometry. For large, high resolution geomodels, subselecting an
area in a memory efficient way is a fundamental operation [Gietzel, 2011]. It is possible to provide
those operations using relational database systems. Storing and accessing the geometry primitives
in table records is becoming computational expensive for very large geomodels. The reason for this
is that beside an overhead of the record access, adjacent information of a geometry primitives is
not used by the database system to optimize the data layout on hard disk for spatial queries. Using
this information it is possible to only access geometry primitives if they are actually required to
process the query.
In this paper we introduce a new technology to store the geometrical primitives in the file system
to provide sub selection queries efficient, while retaining most of the advantages known from systems
based on relational database systems. This system stores data in two different places. Metadata like
name, size, author, used software for model construction, version, raw data connection and other
data provenance information are stored in a relational database system. The large geometry data
is stored outside of the relational database. This can be either a file based system or depending on
custom requirements somewhere else e.g. as binary large object(BLOB) in a traditional database
system, or on any other network storage device. The geometry is divided into spatially grouped
tiles. Those tiles are the basic unit of storing the geometry. This combines the advantages of
storing metadata in traditional environment and large geometry data sets in an optimized way for
commonly used geometrical operations. }
}