Global Positioning System Reference
In-Depth Information
development of such a type of systems can be found in Friis-Christensen et
al. (2007). In this paper the authors describe a use case for the development
of an application that computes forest fi re statistics. By it they present their
architecture for the distributed computation and analyze and discuss some
challenges related to the use of a SOA to perform geographic computation.
The proposed architecture is a traditional multilayer system where the data
access services (WMS, WFS) provide access to data stored in the various
distributed geodata repositories. An additional discovery service layer
is used to provide users with a catalog service to allow them to discover
what data could be useful for their needs. The execution of the diverse
steps for the actual calculation of the various statistics is performed in
the geoprocessing service layer. Finally, the client of the application is a
traditional web application.
The prototype and the underlying architecture are used by the authors
to make some interesting considerations that are worth mentioning here.
First of all, the various steps needed for the effective calculation of the
statistics are performed in a single geoprocessing service. This is a simple
choice and guarantees good performance but limits the general fl exibility
since the different operations required cannot be reused in another
application. A good solution could be the development of a single service
for each functionality, which then have to be chained together in order to
get the desired result. The three types of service chaining identifi ed by the
authors correspond to those supported by the proposed OGC architecture
(transparent, translucent, opaque). In the fi rst type the workfl ow is managed
by a human user, in the second type a service that controls the chain is
invoked and the human is aware of the various steps, in third type an
aggregated service is invoked and the user does not know the single steps.
The fi rst two types of service chaining are analyzed.
The fi rst approach provides a high fl exibility but the client application
must continuously interact with the service and the continuous transmission
of input data can increase the total computation time. The translucent
approach could solve some of the transparent approach issues, since the
entire workfl ow is sent to the service instance in a unique step. However,
the WPS specifi cation would require some adaptation in order to support
such an approach. The second important observation concerns the amount
of time required to perform the processing of voluminous spatial data. This
issue does not only affect the fi nal user experience but might also have an
impact on the services involved in the process. According to the authors
the problem is caused by the synchronous communication mechanism on
which the proposed architecture is based. A possible solution is the use of
asynchronous messaging (supported by the WPS specifi cation) for time-
consuming operations. In such a modality the service response is provided
at a later time in a different communication session (Friis-Christensen et
Search WWH ::
Custom Search