Gridbus Workflow Enactment Engine - Grid Computing: Infrastructure, Service, and Applications - page 129

5.6 Parameterization
Supporting parameterization in the workl ow language is very important
for scientii c applications. It enables scientists to perform experiments
across a range of different parameters without being concerned about the
detailed workl ow description. A parameter dei ned in each task type is
called a local parameter ; when it is dei ned for the entire workl ow, it is called
a global parameter . As shown in Figure 5.9 , multiple parameter types such as
single, range, i le, and enumeration are supported. An example for a single
parameter type and a range parameter type is given in Figure 5.12.
Among these parameter types, range and enumeration types are used
to dei ne a range or a list of parameters valued that the task is required to
be executed. This type of task is called a parameter sweep task [9] and is
structured as a set of multiple execution jobs, each of which is executed
with a distinct set of parameters. Figure 5.13 illustrates parameter sweep
tasks. Two inputs are required by task A and task B ; one is a local param-
eter and the other is a global parameter. The local parameter of A is a
range type parameter whereas the local parameter of B is an enumeration
type. At execution time, the global parameter value is combined with each
local parameter value and generates ten subjobs for task A and three
subjobs for task B .

5.7 I/O Models
As shown in Figure 5.11 , a task receives output data from its parent as its
input through a data link. However, for some tasks, more than one output
could be generated by one output port. For example, such a task could be
a data collecting task that continues to read information from a sensor

<paras>
<para type= “single”>
<name>X</name>
<value type=integer>10</value>
</para>
<para type= “range”>
<name>Y</name>
<min>1</min>
<max>20</max>
<step>2</step>
</para>
</paras>

FIGURE 5.12 Single parameter and range parameter.