Information Technology Reference
In-Depth Information
TRP + US
TRP
D1
s3
D1
s3
D1
s2
D1
s2
US
A1
A2
A3
A4
A5
D1
s1
D1
s0
D1
s4
D1
s5
Fig. 6.
Business process model with data anomalies
anomalies
) in the represented BP model. All these data anomalies cause dead-
locks in the Petri net, so the OLC generated is not complete. For example, in
Figure 7 only the states and transitions outlined with black solid lines are gen-
erated when processing the actual reachability graph that represents the BP
model in Figure 6
5
. States such as
s0
and
s5
are not detected without man-
aging some data anomalies present in the BP model previously. There are two
different groups of anomalous situations, which can be mapped into data-related
problems defined in [2,3]. The resolution of the data anomalies in the BP is out
of the scope of this paper. We explain how to modify the Petri net to solve the
deadlocks and be able to simulate the whole execution of the BP modelled, with
the aim of generating all the states and transitions there represented.
Too restrictive preconditions (TRP).
This kind of problems appear when
the data object specified as input of an activity can be in a different state at
that moment, and so the activity may stay waiting indefinitely. For instance,
in the BP model of Figure 6, activity
A2
will get blocked if
A1
writes data
object
D1
in state
s1
,and
A5
will get blocked if
D1
is in either
s3
or
s4
.
To fix this kind of problems we relax the precondition, assuming that the
state of the data object that caused the deadlock is also a precondition of
the blocked activity.
Unreachable states (US).
In this case, the state specified for the input
data object of an activity is unreachable (either because it does not exist
yet or because the object is in another state at that moment), blocking
the execution of the activity. It appears in activities
A3
and
A5
of the BP
model in Figure 6. To fix it, we assume the data object can be at that state at
that moment and so we continue processing the BP model from the blocked
activity with that “unreachable” state.
There may be other alternatives for dealing with data anomalies, but their study
is out of the scope of this paper. To apply the solutions described above, the
Petri net has to be “re-constructed” in order to obtain a new reachability graph
5
For the sake of clarity we are using a simplified BP to illustrate the data anomalies.
