Information Technology Reference
In-Depth Information
the latter as inputs, this makes the interfaces and thus the synthesis tasks
immediately more complex, which is undesirable with regard to trying to
avoid state explosion problems.
7.2.3
Synthesis
Due to the impact of the characteristics of the domain model (see above),
there is no ultimate synthesis strategy for all applications. Like no other
framework known to the author, PROPHETS enables the user to “play”
with configuration options and synthesis parameters and interactively ex-
plore the solution space. As discussed in greater detail in [233], PROPHETS
provides many configuration options that can be used to tailor the applied
synthesis strategy to the considered application domain. Further on, the work
with the different bioinformatics application scenarios revealed some general
rules-of-thumb for experimenting with synthesis parameters during the ac-
tual workflow design process, especially with search depth and constraints.
In particular, it has turned out to be effective to:
increase the search depth gradually instead of immediately searching for
all solutions, and
identify the required constraints incrementally instead of specifying every-
thing from the beginning.
Initially, the synthesis is free to return all type-correct service compositions
as solutions (from which then one is selected for the actual concretization of
the loose model). This freedom needs to be limited by the user only when
it is necessary, which is in particular the case when the obtained solution
set is of unmanageable size. As such, workflow design in the scope of loose
programming follows a notion of incremental formalization [303], and is hence
in clear contrast to common computer science sense, where typically the
goal is to provide a thorough and complete specification of the application
before it is realized. This is, however, a demanding - if not impossible - task
considering the variant-rich and varying objectives and framework conditions
of the scientific workflow scenarios considered in the scope of this work that
indeed rather ask for an ad-hoc workflow development style. Furthermore,
the target users of the loose programming framework are typically neither
able nor willing to provide thorough specifications in the classical computer
science sense. Thus, the sketched approach to synthesis-supported workflow
design is in fact most appropriate with regard to the envisaged user group
and their applications.
Section 3.3.4 has already illustrated in detail the incremental constraint
specification and search depth exploration for a particular workflow com-
position problem from the phylogenetics analysis workflow scenario. In the
following, these aspects are revisited and regarded from a broader perspective
in order to provide guidelines that are generally applicable.
 
Search WWH ::




Custom Search