Information Technology Reference
In-Depth Information
appropriate behavioural abstraction of processes. Throughout this paper, we
will demonstrate that the essential variant management questions can be an-
swered based on set algebraic operations of complementation, intersection, and
union, and the relations of set equivalence and inclusion. Although there are sev-
eral works on particular subsets of these aspects such as inclusion (or behaviour
inheritance) [7,8,9] and union (or merging behaviour) [10,11,12,13], we currently
miss an overarching set algebra which provides the means to eciently calculate
with behaviour.
There is a wide spectrum of behavioural abstraction available upon which alge-
braic operations could be defined. Existing work on behaviour inheritance [8,9,14]
builds on an adapted notion of branching bisimilarity. That is, activities that are
without counterpart can either be
blocked
or
hidden
when assessing behavioural
equivalence. Such an approach has the drawback that the underlying notion of
branching bisimilarity is computationally hard as it is based on state space anal-
ysis, cf., [15]. Therefore, we base our set algebra on behavioural profiles, a be-
havioural abstraction of process models that can be computed eciently for a
broad class of models. Behavioural profiles capture the essential behaviour of the
set of traces of a process in terms of order constraints, and they are insensitive to
skipping of an activity. Differences related to causal constraints are often observed
among process variants, and behavioural profiles have been found to provide a suit-
able abstraction to reason about consistency of process variants [16].
Our contribution is a formal definition of a set algebra for calculating with
behavioural profiles of process variants. Our approach is inspired by work on a
set algebra for service interaction [17]. The concepts are applied to an industry
model collection, for which we identify variant clusters and analyse behavioural
commonalities. In this way, our work informs formal research on behavioural
inheritance as much as engineering approaches towards merging of behaviour.
The remainder of this paper is structured as follows. The next section illustrate
the use case in more detail and presents formal preliminaries. Section 3 defines
a set algebra for behavioural profiles, which is applied in Section 4 to address
various questions on managing process variability. We present findings from a
case study with industry models in Section 5. Finally, Section 6 discusses our
approach in the light of related work, before Section 7 concludes the paper.
2 Background
This section introduces the background of our work. First, Section 2.1 elaborates
on the use case of managing decoupled process variants. Subsequently, Section 2.2
presents our formal model and Section 2.3 defines behavioural profiles.
2.1 Challenges for Managing Decoupled Process Variants
We illustrate our use case using the two process models depicted in Fig. 1. Both
models capture the process of registering a newborn and are adapted versions of
the models obtained in different Dutch municipalities, cf., [18]. The administra-
tive processes of Dutch municipalities are rather similar due to legal regulations
