Conclusion - User-Level Workflow Design: A Bioinformatics Perspective - page 183

Fig. 9.1 Trade-off between simplicity and generality in software environments

design takes place on the level of abstraction defined by the framework,
rather than on the user level.

• Domain-specific workflow systems focus on user-level service composition

making use of a specific (semantic) domain model, which allows the user
to work with the services and workflows using the technical language of
his domain. These frameworks often also apply semantics-based methods
for (semi-) automatic workflow composition. They are, however, usually
restricted to the services covered by the domain model.

According to this classification scheme, Bio-jETI is both a domain-
independent and a domain-specific workflow management system: The jABC
framework with its broad range of established plugins provides a mature
(domain-independent) workflow management system, and the loose program-
ming approach implemented by the PROPHETS plugin allows for the in-
corporation of (domain-specific) semantic support on demand. In fact, the
workflow development styles complement each other rather than having to
be applied mutually exclusive. Thus, Bio-jETI provides a maximum balance
of simplicity and generality. This acknowledges the observation that conve-
nience for the user can not be achieved without specializing the software, but
at the same time shows that domain-specialization does not have to impact
the flexibility that is supported by a workflow framework.
In greater detail, Bio-jETI is a framework for service-oriented, model-
driven design, execution, verification and deployment of bioinformatics work-
flows, extended by functionality for semantics-based, (semi-) automatic work-
flow composition. It is unique in its application of different flavors of formal
methods, which facilitate constraint-guarded and constraint-driven workflow
design:

• Constraint-guarded workflow design means that workflow design is accom-

panied by the continuous evaluation of workflow-level constraints, alert-
ing the workflow designer when constraints are violated. In Bio-jETI, the
GEAR model checking plugin is used to monitor workflow design in terms
of constraints that are expressed in a temporal logic. Thus, in addition
to the speedup of in silico experimentation and the abstraction from
programming details that are also achieved by other commonly known
systems for bioinformatics workflow design, constraint-guarded workflow