Martin Brändli
and Andreas Ernst
Spatial Data Handling
Division
Department of Geography
University of Zurich
Switzerland
brand@geo.unizh.ch
aernst@geo.unizh.ch
Different concepts
and developments of managing persistent data (flat files, relational and
object-oriented databases) have led to significant obstacles for the development
of interoperable systems. This fact is especially true for the domain of
geographical information systems (GIS). Currently available GIS are very
similar with respect to their architecture and functionality. Nevertheless,
they are not compatible with each other concerning the data they manage.
Most systems distinguish between the management of geometrical and descriptive
data. For performance reasons, geographical data are mainly stored in proprietary
databases. Consequently, the interoperation between different GIS is very
difficult, since many different interfaces have to be supported. FRIEND
(Framework for the integration of environmental and geographical data)
is an interdisciplinary project in the area of GIS. Its main objective
is the integration of heterogeneous, space-related data repositories under
a common roof. In particular, it aims at solving (or at least decreasing)
the integration problem posed by the necessity to develop and maintain
many interfaces. This paper intents to outline the integration approach
which is used in FRIEND. Starting from a framework that approaches data
integration from a database technology view in general, it specifically
describes the integration of geodata components using an approach based
on OGIS (Open Geodata Interoperability Specification), implemented with
the Java programming language.
Typically, there
are four choices to bring together data from different sources: migration
(physical transfer of data from different systems to one new system), data
transfer between two systems, use of common data catalogs, and federated
database management systems (FDBMS). FRIEND is based on the last option.
It allows the integration of data without migration and is thus called
"logical integration". Within a FDBMS, local database management systems
(components of the federation) keep their autonomy. Some of the issues
of the architectural design of this logical integration will be discussed
throughout the paper. FRIEND aims at developing a generic solution for
the integration problem. This generic solution is implemented with a framework
that is adjustable to particular situations. It utilizes a set of interoperable
objects following an object-oriented approach, and is implemented as a
layer connected to an object-oriented database management system which
handles data that have to be stored on the global level of the federation.
The data model used in the framework is an ODMG-compatible (Oject Database
Management Group) object-oriented model.
In order to integrate
GIS-components on the logical level, the object-oriented data model has
to be extended by geographical data types and methods. Since one of the
goals of the FRIEND-project is to reach high conformity with international
standards, the integration of and access to GIS-components is based on
OGIS interfaces. As a consequence, the object-oriented data model used
for the integration layer has to be completed with OGIS-compliant data
types and concepts. Even if the integration layer is developed as generic
as possible, we start the implementation of the integration of GIS-components
and geographical data from a real world integration problem which exists
at the municipality of the city of Zurich (Switzerland), and which may
be similar or comparable to many other urban administrations. During the
past thirty years the different institutions of the municipality (surveyors's
office, water and energy supply utilities, etc.) developed their own spatial
data handling solutions leading to a heterogeneity of systems and data
models in conjunction with multiple and inconsistent data collections of
the same objects. Today, this situation is not acceptable anymore and the
joint use of the data is required. The most important demand on data integration
concern the supply utilities in conjunction with the surveyors's office,
since the supply utilities heavily depend on consistent survey data. The
complexity of the present data as well as the apprehension of loosing autonomy
make a real world integration application difficult at the time being.
However, the knowledge of the situation acquired by an analysis of the
current state of the involved institutions motivated us to implement a
so-called "miniworld", which simulates the actual situation by substantially
reducing the complexity and size of the data and the number of involved
institutions.
This miniworld
is designed in the following manner: The federated system consists of two
geodata components including the surveyors's office exposing objects such
as parcels, buildings, and landmarks on the one hand, and the water supply
utility exposing features such as reservoirs, pipes, pumps, consumer sockets,
and sleeves on the other hand. In order to install a heterogeneous environment,
the surveyors office is implemented with an object-oriented, the water
supply utility with a relational database system. The realization of the
miniworld involves four steps, each of it coming closer to the real world
situation. It corresponds to the five-level schema to describe the architecture
of a FDBMS proposed by Sheth and Larson (1990):