Database Reference
In-Depth Information
While provenance information is typically used “post-mortem,” that is, af-
ter a workflow is run to interpret, validate, or debug results, it can also support
runtime execution monitoring.
71
,
75
The SDM dashboard application supports
runtime execution monitoring using the architecture of Figure 13.5. The dash-
board is illustrated in Figure 13.6, which shows the on-the-fly visualizations
generated by the monitoring workflow described in Section 13.3. Dashboards
generally display condensed information about the status of workflow pro-
cesses, data, the execution environment, and so forth. In addition to providing
a high-level overview, such dashboards may also offer a way to navigate into
details of runtime progress and provenance trace information, and to show
trends in the output data or execution performance.
Other scientific workflow systems have similar capabilities to those de-
scribed above. For example, Pegasus has been integrated with the PASOA
provenance system.
76
Within Triana, provenance information can include the
components executed, their parameters, and the datasets that pass through
during execution. A data provenance system for Taverna is described in
Missier et al.
63
13.6 Workflow Sharing and myExperiment
Understanding the whole lifecycle of workflow design, prototyping, pro-
duction, management, publication, and discovery is fundamental to devel-
oping systems that support the scientist's work and not just the work-
flow's execution. Supporting that lifecycle can be the factor that means a
workflow approach is adopted or not. Workflow descriptions are not sim-
ply digital data objects like many other assets of e-Science, but rather
they capture pieces of the scientific process: They are valuable knowledge
assets in their own right, capturing valuable know-how that is otherwise often
tacit. We can conceive of packages of workflows for certain topics, and of work-
flow “pattern books,” that is, new structures above the level of the individual
workflow. Workflows themselves can be the subject of peer review, and can
support reproducibility in the scholarly knowledge cycle. We can view them as
commodities, as valuable first-class assets in their own right, to be pooled and
shared, traded and reused, within communities and across communities. This
perspective of the interacting data, services, workflows and their metadata
within a scientific environment is a
workflow ecosystem
. Understanding and
enabling this ecosystem is the key to unlocking the broader scientific potential
of workflow systems.
