Information Technology Reference
In-Depth Information
Requirement 5.2: runtime behavior
Simulation and debugging functionality enables the detection and analysis
of workflow modeling errors that become obvious during execution. Ideally
the workflow system should also keep track of the workflow executions in
a kind of event history, logging which workflows were executed when and
with which result. This information can then be used for the analysis of the
workflows in terms of, for instance, load factor or execution time statistics,
as well as frequently occurring errors or performance issues [166].
Requirement 6: workflow execution
Execution is of course central for scientific workflow applications [242, 128].
Therefore the workflow model should not only represent the underlying pro-
cess, but rather contain also all information that is necessary for its execution,
for instance by using an associated interpreter or a specific execution engine.
Such information may be provided at the level of the process building blocks
or, globally, at the workflow model level.
Generally, workflow execution can take place within or outside of the work-
flow environment, and ideally both possibilities should be supported:
Requirement 6.1: execution within the workflow environment
Execution within the workflow system is convenient when the workflow
developer is also the user of the workflow, as he can continue to work in
the familiar environment. Furthermore, runtime validation becomes eas-
ily possible (see above). Ideally, ad-hoc execution of the workflow model
should be carried out by an interpreter component that is provided by
the workflow environment.
Requirement 6.2: execution outside of the workflow environment
Execution outside of the workflow system is required in case that accept-
able computation times can only be obtained when the workflow is exe-
cuted, for instance, on a grid or cluster infrastructure [166]. This requires
the deployment of the workflow to, for instance, a workflow engine or the
compilation of the workflow model into a stand-alone application, which
is ready to be used by others. Therefore code generation and application
deployment functionality should be provided by the workflow modeling
framework.
This list of requirements is in fact rather general and as such applies to
different scientific application domains. More specific requirements, especially
those related to supported data formats and styles, are subject to the nature
and complexity of the concrete application for which the system is to be
used. In fact, many requirements for workflow systems that are identified
in the general computer science or business processes context (such as work
lists, role assignments, time-based triggers, or deadlines) are considered less
relevant for the scientific domain [166].
 
Search WWH ::




Custom Search