Information Technology Reference
In-Depth Information
is inherent in the SLGs, which are control-flow representations of workflows,
while data handling is done via the execution contexts. In fact, it is part of
the jABC philosophy to hide the data (flow) to a certain extent behind the
behavioral workflow model: the control-flow is represented by the graphical
workflow model, while the data-flow is defined via component configurations.
Thus, data dependencies between the single services do not obstruct the work-
flow representation, and even large workflow with complex data flows are still
easily readable. In contrast, the data-flow approach that is taken by the ma-
jority of popular bioinformatics workflow systems is based on the philosophy
that the data is central to the analysis workflows and have thus to be put
in the foreground of the workflow definition. However, as Section 8.2 will
discuss in detail, control-flow-oriented workflow models like Bio-jETI's sup-
port the definition of workflows with greater computational complexity. At
the end of the day it depends both on the concrete application as well as on
the preferences of the workflow modeler which data handling concept is more
appropriate.
Most systems support hierarchical workflow models that allow for encap-
sulation of sub-workflows into reusable components. In Bio-jETI, complex
workflows are enabled by the hierarchy concept of the SLGs.
Domain Modeling (Requirement 3)
The vast majority of systems share the “open world service assumption” [242]
and allow for the extension of the component libraries. However, support for
(semi-) automatic integration and service management is realized in very
different ways.
In Bio-jETI, the open world assumption for services is met by the tech-
nical concepts of the SIBs (freely extensible Java classes), and additionally
supported by the jETI platform and the SIBCreator plugin. Service manage-
ment via the SIB browser is supported by the Taxonomy Editor plugin, while
additionally the PROPHETS plugin provides the means for semantic service
management and transparent discovery.
Most systems provide no means for the semantic enrichment of the domain
model at all. A couple of systems support basic semantic domain modeling,
but rely on predefined domain models that are only supposed to be changed
by their developers, that is, they use the knowledge that is provided by the
service and data type descriptions and ontological classifications, whereas all
additional domain knowledge (if any) is hidden in specifically designed com-
position algorithms. Wings defines the creation of the workflow templates
as distinct phase, which is part of the domain modeling together with the
component and constraint specifications. GNU make, SADI and the Seman-
tic Workflow Tool can deal with changing domain models, but they do not
provide explicit support for domain modeling.
Bio-jETI is the only system among the presented that offers comprehensive
semantic domain modeling support (including the both definition of domain
Search WWH ::
Custom Search