Information Technology Reference
In-Depth Information
tools referenced in the model are retrieved from the Tool Manager and the
needed service is invoked through the Connector, (ii) input data referenced in
the model is downloaded automatically from the repository and provided to
the service, and (iii) output data is uploaded to the repository. The User In-
terface contains the image of the platform independent version of the process
model, where the state of the execution and the related data can be traced.
3
Implementation of the Tool Integration Framework
The tool integration framework is implemented as an Eclipse based tool and
the standard Eclipse extension mechanism is used for the extendability of the
framework.
Tool Management.
The Tool Manager component was developed in the Sen-
soria project and it is called SDE [7]. This provides a
tool extension point
,
to where the tool Connectors shall register. A local tool repository is com-
posed of this registered tools. If the Tool Manager is deployed on multiple
hosts, it is possible to add the remote tool repositories to the local one and
access the tools registered there through r-OSGi communication. This way
the framework can be used in a distributed environment.
Data Management.
The general component of the Artefact Manager provides
a
repository extension point
, to where the repository connectors can register.
Currently connectors are implemented for the file system (as this is handled
also as a data repository), for the Java Content Repositories (the Apache
Jackrabbit reference implementation is used [11]), and for the Subversion
repository [12].
Process Modeling.
For the creation of process models an Eclipse GMF [13]
based graphical editor is developed. Here the process can be composed of the
services deployed on the local or on a remote tool repository, and artefacts
contained by the available data repositories can be referenced.
Process Execution.
The JBoss jBPM workflow engine [14] is integrated as the
process execution component of the framework. It provides a persistent store
for the processes, where information about all executions are stored. The
process execution user interface part of the framework provides a catalogue
of the deployed and instantiated processes, makes possible to deploy and
instantiate new processes and to trace the state of the selected process.
jPDL (jBPM Process Definition Language, [15]) is the native process for-
mat of jBPM, this platform specific process model is transformed automat-
ically from the “general” process model created in the process editor. We
have extended the jPDL language with the notion of data nodes, and pro-
vided action handlers for the tool and data nodes. The former invokes the
referenced service on the local or remote host, while the later transparently
downloads input data of the tools from repositories or uploads output data
to the repositories as determined by the process model.
