Global Positioning System Reference
In-Depth Information
Scientific workflow systems and component-based modeling
frameworks (Argent 2004; Argent et al. 2006) share some commonalities. Both
typically support datafl ow-oriented workfl ows that can be computationally
expensive, thereby data-intensive use-case scenarios characterize both sorts
of systems.
Ludäscher and Goble (2005) however highlight annotation as the
differentiator element, i.e., the ability to document a workfl ow in form
of descriptions of the results of the workfl ow execution, inputs data sets
used, restrictions and post-conditions after execution, and so on. Scientifi c
workfl ows are often more annotation-intensive because by defi nition they
pursue reproducibility (Jasny et al. 2011; Mesirov 2010) so that an annotated
workfl ow with useful information can be replicated and reproduced in other
scenarios later on. Reproducibility thus requires detailed context metadata
and data provenance information (Cohen-Boulakia and Leser 2011). For
instance, a common pattern in environmental sciences is to perform the
same set of analytical tasks several times but changing the input data sets
in each run. Take the example of a watershed runoff model: while the
logic of the model itself remains invariable, the results change because the
model is fed with distinct time-series data sets in each run. That is, control
fl ow defi nitions are not as relevant and critical as data sets used in each
execution. In these cases, annotation techniques (e.g., context metadata,
provenance, traceability) allow modelers to document properly scientifi c
workfl ows along with data sets used and produced, and to create accurate
records of scientifi c experiments and procedures in order to be reproduced
or analyzed later. The immediate benefi t from workfl ow annotations is
the ability to reuse others' workfl ows in a confi dent and reliable way, as
scientists can reproduce and compare same scientifi c workfl ows applied to
distinct use-case scenarios, leading to more transparent and reproducible
science (Mesirov 2010).
In the workfl ow literature there is not a unique solution but dozens of
scientifi c workfl ow systems specialized in different scientifi c domains and
disciplines (Yu and Buyya 2005; Deelman et al. 2009). This vast array of
systems suggests that there is no a single workfl ow system to handle with
the diversity of scientifi c workfl ows, nor is there a common data model
for all scientifi c workfl ows. All of this implies a lack of interoperability
between workfl ow management systems (Elmroth et al. 2010). Given that,
we focus in this section on widely-used scientifi c workfl ow systems and
particularly in their support for web service technologies. Recent reviews
on scientifi c workfl ow systems (Yu and Buyya 2005; Rahman et al. 2011) also
include the scientifi c workfl ow systems analyzed below but complemented
here by having a close look at the interplay between geo-computation and
geoprocesing web services capabilities.
Search WWH ::
Custom Search