Information Technology Reference
In-Depth Information
comparison between them and BPEL. Instead, we pick up a represent-
ative, i.e., Taverna that is used by the aforementioned caGrid project,
and compare it with BPEL.
Developed in the UK by the myGrid consortium, Taverna is an
open-source workbench for the design and execution of scientific
workflows. Aiming primarily at the life science community, its main
goal is to make the design and execution of workflows accessible to
researchers who are not necessarily experts in Web services and
programming. A Taverna workflow consists of a graph of activities
that represent Web services or other executable components. Activities
receive their inputs from input ports and produce outputs to output
ports. An arc in the graph connects an output port of one activity to an
input port of another, and thus, it specifies a data dependency. Work-
flows are specified using a visual language and executed according to a
dataflow computation model [182]. Taverna also provides a plug-in
architecture such that additional applications, for example, caGrid
infrastructure management and services, can be populated to it, as
discussed later in Section 7.2.2
caGrid Workflow Toolkit
.
We choose BPEL and Taverna as representatives for business and
scientific workflow languages, respectively, for the following reasons.
Firstly, they are typical and widely used in business and scientific
domains, respectively. BPEL is the
de facto
standard for business
workflows and adopted by IBM, Microsoft, Oracle, SAP, and so on.
On the other hand, Taverna is used by many organizations and in various
areas including life science, astronomy, and chemistry. Secondly,
Taverna and BPEL have a historical connection. Taverna originally
supported a subset of WSFL [183] which is an ancestor of BPEL.
Therefore, from this comparison, we can find out how two languages
from the same origin evolve to serve different purposes. Last but not
least, in caGrid, we use both BPEL and Taverna. Thus, we have gained
sufficient first-hand experience in comparing them.
Although BPEL and Taverna share some common modeling
elements, these come as a result of different design motivations,
which account for different programming styles. BPEL is designed for
the interactions among business partners that expose their functions
via service interfaces and for facilitating their
interoperability
.The
major motivation for the design of Taverna, on the other hand, is to
streamline
the
dataflow
among services,
scripts,
and other
applications.
