Information Technology Reference
In-Depth Information
[184]. This comparison reveals that scientific workflows (using Taverna
as an example) do have unique characteristics compared to their
business counterparts, thereby deserving special attention.
Our experience in using both Taverna and BPEL as the candidate
solutions for caGrid illustrates that:
BPEL offers a comprehensive set of primitives to model work-
flows of all flavors (control-flow oriented, dataflow oriented, and
event-driven), with full features (process logic, data manipula-
tion, event and message processing, fault handling, etc.). As an
imperative language, BPEL offers fine control and close com-
munication to workflow engines. BPEL is capable of handling
dataflows, although in some cases, the modeling experience is
cumbersome and tedious.
Taverna provides a compact set of primitives that eases the
modeling of dataflows. Its functional programming framework
makes many routine tasks invisible to the programmer such that
users can focus on specifying “what to do” instead of “how to do
them.” The workflow engine handles routine tasks such as
variable initialization, parallel execution through implicit itera-
tion, and pipelined execution.
After a pilot study on both Taverna and BPEL, caGrid decided to
use Taverna to orchestrate its services, thanks to its aforementioned
advantages in dataflow modeling. Despite this fact, no rational user will
argue against the superiority of BPEL for certain tasks. Different tasks
will result in demand for different types of workflow languages and
systems.
7.2.2 caGrid Workflow Toolkit
Motivation
caGrid project selected Taverna as its workflow execution system of
choice due to its integration with Web service technology and support
for a wide range of Web services, plug-in architecture to cater for easy
integration of third-party extensions, and a broad user base within the
bioinformatics/biomedicine community [185]. The caGrid Workflow
Toolkit, an extension to the Taverna workflow system, is designed and
