Information Technology Reference
In-Depth Information
service integration for workflow applications in general, and for semantic do-
main modeling and synthesis application the scope of the loose programming
framework in particular.
7.1
Synthesis Evaluation
Before starting with the details of the evaluation, the following briefly summa-
rizes and compares the four considered application scenarios with particular
regard to the constraint-driven workflow design with PROPHETS:
Application Scenario 1: Phylogenetic Analysis Workflows
The EMBOSS domain model comprises the largest amount of services (more
than 430) among the four examples. It has been automatically generated
from the ACD files provided with the EMBOSS release, which also contain
semantic annotations in terms of the EDAM ontology. This application is
particularly challenging for the synthesis method, since the domain contains
many services for similar purposes (sequence analysis) that also work on simi-
lar data types, so that the synthesis can easily find extremely many solutions
already in relatively small search depths. Consequently, the provisioning of
adequate domain-specific and problem-specific constraints is essential in or-
der to arrive at a manageable amount of solutions.
Application Scenario 2: GeneFisher-P
The GeneFisher-P domain model also applies the EDAM terminology, but the
service annotations themselves have been defined manually. It is much smaller
than the EMBOSS domain (only comprising 10 services) and also much more
specific (tailored to primer design workflows). Hence considerably less work-
flows are possible based on the GeneFisher-P domain model and thus the set
of solutions that is obtained by the synthesis is more or less manageable also
without defining comprehensive sets of additional constraints. The exemplary
workflow composition problem also illustrates the over-constraining of speci-
fications and some of the problems that can be caused by imprecise service
interface descriptions.
Application Scenario 3: FiatFlux-P
The FiatFlux-P domain model does not apply EDAM terminology, since the
required terms are not (yet) covered by the ontology. Accordingly, service
annotations and taxonomies have been created manually and tailored to the
specific application. The FiatFlux-P domain is only slightly larger than the
GeneFisher-P domain (12 services), so that the number of solutions that
Search WWH ::
Custom Search