Information Technology Reference
In-Depth Information
In the second scenario, a workflow that embeds the best practice
of one experiment can be shared with the community. For
example, a lymphoma diagnosis workflow built and shared by
one researcher can be used by another for the purpose of breast
cancer analysis.
Taverna versus BPEL: Not an Old Wine in New Bottle
Currently, there are many scientific workflow systems available and the
well-edited topic [72] provides a summary of them. Among them, many
systems are designed for either only composing Web services or
providing support to compose services. They include Askalon [176],
Kepler [177], GPEL [178], OMII-UK [179], Taverna [180], Triana
[181], and Trident [163]. Each of them provides a graph-based interface
for service composition, with an underlying workflow metamodel. The
workflow metamodels used by these service-based systems are either
adopted from industry standard or homegrown.
GPEL and OMII-UK adopt the industry standard languages WS-
BPEL. WS-BPEL is an XML-based specification that describes the
behavior of a business process composed of Web services, and the
workflow itself is also exposed as a Web service. Originally designed
for business workflows, BPEL is also embraced by the scientific
community. Obviously, adopting industry standard like BPEL can bring
some advantages such as rigorously defined model syntax and seman-
tics, readily available software tools, and portability of workflow
specifications. However, a distinction can be made between workflow
languages designed for business processes and those for scientific
pipelines. This distinction arises from the different nature of business
and scientific applications. Business workflows present the routines
inside or between enterprises, and therefore, comprehensive control-
flows embodying business rules, process integrity (including transac-
tion and compliance), and human involvement are the major concerns.
Scientific workflows present the experiments conducted by scientists,
and therefore, fast data processing and scheduling of computing
resources are the major concerns.
To deal with the unique features of scientific workflows, Askalon,
Kepler, Taverna, Triana, and Trident have devised their own workflow
metamodels. Due to the diversified and specific concerns addressed by
these homegrown models, it is difficult to provide a comprehensive
