Introduction - User-Level Workflow Design: A Bioinformatics Perspective - page 6

conception as well as the technical realization of services depends on the
characteristics and requirements of the specific application scenario.
For the image processing workflow example of Figure 1.3, for instance, the
ImageMagick [310, 10] command line tools can be used as basic services. Its
flexibly parameterizable compare , composite , convert , identify , mogrify
and montage tools provide a comprehensive collection of ready-to-use im-
age processing functionality that, in contrast to common image processing
software, runs autonomously without a graphical user interface. Technically,
these tools can either be executed directly via the command line interface, or
encapsulated into some wrapper script or the like that facilitates integration
with other software environments.
Workflow applications regard services as first-class citizens that provide
clearly defined, user-accessible units of functionality [197]. This service-
orientation clearly facilitates the domain-specificity of the workflow applica-
tions, and furthermore inherently supports virtualization and loose coupling
of the workflow building blocks [206].

Model-Based Workflow Development

In model-based workflow development frameworks, users can easily build and
handle workflows without knowledge of details of the underlying technologies
(cf., e.g., [208]). The level of abstraction that is necessary in this respect is
essentially achieved by:

• Abstraction from technical details of the single services through virtual-

ization concepts that enable the integration of heterogeneous resources
into a homogeneous environment.

• Abstraction from syntactical details of a particular (programming) lan-

guage through the application of a graphical and thus more intuitive for-
malism.

The first kind of abstraction goes in line with the service-oriented nature of
workflows, and applications profit from the existing (web) service infrastruc-
ture of many application domains. Accordingly the loose coupling of services
and thus a lightweight workflow composition [205] becomes possible. The sec-
ond kind of abstraction allows the user to focus on the workflow aspects of the
application (such as the execution order of services, or transition conditions).
The common graphical workflow models are intuitive programmatic realiza-
tions of the workflows, directly providing comprehensive user-level documen-
tation of the experimental procedures. They also enable workflow systems to
support the adding, removing, or exchanging of services graphically, as well
as changes to the control or data flow. In such environments, the agility in the
development and the quality and maintainability of the workflow applications
are considerably higher than with conventional implementation.
Figure 1.4 shows an exemplary graphical workflow model for the example
workflow depicted in Figure 1.3 that makes use of ImageMagick services as