Introduction - User-Level Workflow Design: A Bioinformatics Perspective - page 10

Fig. 1.5 Scientific workflow life cycle (following and extending [193])

interpretation of workflow results. Based on these analyses researchers decide
if they have to adapt the experimental setup and to go through the workflow
cycle again in order to improve their results. Workflow management systems
are typically primarily concerned with the workflow design, preparation, and
execution phases.
As Figure 1.5 also shows, the initial hypothesis formulation and the anal-
ysis of the results take place at the user level, the workflow preparation and
execution phases are located at the technology level, while the workflow de-
sign phase is concerned with both user-level and technology-level aspects.
Accordingly, simplification of the workflow design phase can be addressed in
two complementary directions:

• “upwards”, that is, in terms of leveraging workflow design to an even

higher and more conceptual level, thus bringing it closer towards the user.
This goes in line with the (visions for) workflow systems supporting se-
mantic service composition as described in, for instance, [66, 192, 256,
122, 121, 257, 120, 110, 68, 258].

•

“downwards”, that is, with respect to achieving a better integration with
the underlying technologies, such as grid, cluster and cloud infrastructures,
based on the argument that the processing of large amounts of data is
crucial (cf., e.g., [27, 355]).

Note that these complementary directions of workflow development do not
necessarily have to contradict each other, but can rather complement each
other due to a clear separation of concerns: instead of having a single workflow
model trying to cover both user and execution logic, it should finally be pos-
sible to have an (abstract) model in terms of the user's domain language,
which can be automatically translated into highly specialized execution
models.