Support for Cooperative Experiments in e-Science: From Scientific Work flows to Knowledge Sharing - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 140

covers several well-known systems, namely ASKALON (Fahringer et al. 2005 ; Qin
et al. 2006 ; Deelman et al. 2009 ), GridNexus (Brown et al. 2005 ) , Grid Work fl ow
Execution Service (GWES) (Hoheisel 2006a; Bubak and Unger 2006 ) , ICENI
(Mayer et al. 2006 ), Karajan (von Laszewski and Hategan 2005 ) , Kepler (Ludäscher
et al. 2009 ; Altintas et al. 2004 ; Buyya et al. 2000 ), Pegasus (Callaghan et al. 2009 ;
Deelman et al. 2009 ), Taverna (Oinn et al. 2002 ) and Triana (Taylor 2006 ; Taylor
et al. 2005 ; Harrison et al. 2008 ; Churches et al. 2006 ). We review theses systems in
light of their Design features, Execution support and Support for Interoperability,
and Results dissemination and information management support .

7.4.1 Design Features

Since workflows are intuitively depicted as graphs, it is no surprise that most
SWMS offer a graphical workflow composer for building such graphs. Although
there seems to be a consensus as to the basic notion for a workflow vertex being
a computation task and an edge being a data and/or control flow, different SWMS
tend to differ on the actual graph modeling. Common differences are cyclic versus
acyclic and stateful versus stateless graph models. SWMSs such as ASKALON,
Taverna, Karajan model workflows based on Directed Acyclic Graphs (DAG).
Since pure DAGs do not model conditional branches and loops, such systems
augment the DAG model to include these primitive control structures. ASKALON
and Karajan also include advanced control structures such as parallel constructs.
Some systems such as GWES do away with DAG modeling altogether and instead
use Petri nets. Petri nets differ from DAGs by modeling control flow and, most
importantly, model workflow state through the use of token transitions.
Furthermore, Petri nets well understood properties such as deadlock and conflict
can aid in model analyses and optimization. Kepler goes a step further and allows
different types of models, which it achieves through directors. This notion of dif-
ferent Models of Computation (MoC) allows for greater flexibility as many other
SWMS only allow one MoC. Common MoC in Kepler are: Process Network,
Dataflow, Discrete Events, Synchronous/Reactive . It is most often the case that a
graphical workflow composition is synthesized to an XML-based language,
which facilitates sharing and reusability. Such languages include AWGL, JXPL,
GWorkflowDL and Scufl used by ASKALON, GridNexus, GWES and Taverna
respectively.
Scientific workflows are collaborative in nature and hence workflow sharing
becomes an important feature in an SWMS. Such collaborative features are included
in Kepler and Taverna, which support semantic queries for components that can be
reused in new workflows. Taverna also includes a browser which provides naviga-
tional capabilities over data stored in the myGrid (Stevens et al. 2003 ) information
repository (MIR), which include experimental designs, experiment results, and
intermediate data.